Their goal was simple: generate qualified leads from both traditional SEO and AI search. 

To achieve this, we: 

• Applied Pain Point SEO to rank for high-buying-intent keywords in Google
• Built a GEO Topic Map that teaches LLMs when to recommend them in AI search

So whether prospects search on Google or ask LLMs (like ChatGPT or Claude) about survey software, they have a strong chance of being recommended as a solution.

Below, we explain the principles behind our GEO and SEO strategy and share examples of our results.

Our Strategy for Gaining Both Traditional SEO & AI Search Visibility for Our Client

The core principle of Pain Point SEO is prioritizing high-buying-intent keywords over the top-of-funnel informational content most brands focus on. Our data shows these bottom-of-funnel keywords convert 10x to 25x better than low-intent queries.

For example, instead of targeting keywords like “What is a net promoter score” or “What is a focus group,” where the reader still has a lot of learning to do before buying, we prioritize keywords like “best survey software for enterprises” or “best customer research tools.”

The same principles apply to AI search visibility, since LLMs rely on web results when users ask for product or service recommendations. But there’s an important distinction.

Unlike Google, where people type short, predictable queries, LLM conversations are longer, more detailed, and often informed by prior interactions. In many cases, these systems have weeks or months of context about a user, which they factor into their recommendations.

As a result, LLM recommendations are more personalized than Google search results.

Take two users evaluating our client’s survey software: a startup running feedback loops with early users and an enterprise looking for affordable scale. If both search “best survey software” on Google, they’ll see similar results. But the same query in an LLM would factor in previous conversations, and the recommendations would differ based on each user’s context.

In AI search, it’s less about ranking for well-defined keywords and more about being a brand the model recommends when a relevant topic comes up. 

That’s a fundamentally different challenge than ranking for short keywords. LLMs need enough context to make a compelling case for your product — and that context has to be relevant to each user’s situation.

We call this approach Topic-Based GEO.

Topic-Based GEO argues that to maximize AI search visibility, you need published content that gives LLMs what they need to position your product as the solution to specific use cases and situations (“topics”).

We visualize the entire universe of topics relevant to your business as a map and call it your GEO Topic Map. 

To build it, we start by brainstorming every use case and pain point someone might ask an LLM about survey software.

The result is our Topic Map: a body of content that addresses every conversation or situation in which an LLM might reasonably recommend our client’s survey software.

For example, our client works with:

• Startups using survey software to gather early customer feedback
• Enterprises with strict security requirements
• DIY market researchers looking for affordable tools

We then created highly specific content for each of these use cases. Below, we share our results.

Examples of Keywords We’re Ranking For & Our AI Search Visibility

Example 1: “DIY market research tools”

One of the first articles we published for our client targets conversations where a user is asking about DIY market research tools, a common use case among their ideal customers.

If we look inside Traqer.ai — the tool we developed to track AI search visibility — we’re showing up in the top 5 positions in ChatGPT, Perplexity, and Google AIO for the topic “DIY market research tools.”

In Traqer, we also see that whenever someone enters a prompt about DIY tools for market research, our article is cited more than any other.

As a bonus, we’re ranking 4th for “DIY market research tools” on Google and generating leads from organic search.

Example 2: “CPG market research companies”

We also want LLMs to recommend our client’s survey software when someone asks for CPG market research companies, since CPG (consumer packaged goods) brands are among the heaviest users of market research.

So, we published a highly differentiated article on “CPG market research companies” that shows how our client’s survey software addresses the specific pain points of CPG brands.

In Traqer, we can see that we’re in the top 3 recommendations in ChatGPT, Perplexity, and Google AIO for prompts related to CPG market research companies.

LLMs also pull from our article more than any other piece on this topic.

We’re also currently ranking 1st for the keyword “CPG market research companies” on traditional Google search.

Example 3: “Enterprise survey software”

A large percentage of our client’s customer base is enterprise clients looking for software that scales well and doesn’t become increasingly expensive as they add users. So, we knew that enterprise survey software is an important topic to show up for on LLMs.

In Traqer, you can see that we consistently rank in the top 5 for queries related to enterprise survey software.

Our article is also the source AI cites most often when users ask about enterprise survey software.

On top of that, the article we published targeting this keyword is ranking 2nd on Google.

These are just a few examples, since we can’t realistically cover every topic we have visibility for in AI search.

Overall, the topic map we built gives our client AI search visibility across 100+ high-buying-intent topics.

Our Client’s AI Search Visibility Versus Competitors

We also think it’s worth comparing our number of AI citations with those of competitors, as this helps put the effectiveness of our AI visibility strategy into context.

As shown in the screenshot below, our article is cited nearly twice as often as SurveyMonkey, Involve.me, and Quantilope on the topic of “AI survey tools.”

On the topic of “enterprise survey software,” AI tools are citing our client more than competitors like SurveyLab and Zonka Feedback.

The same is true for “conjoint analysis software.” We have more citations than Conjointly, OpinionX, and Qualtrics.

These are just a few examples, but the pattern holds across most topics. For most high-buying-intent prompts, AI tools cite our articles more frequently than those of our competitors.

Key Takeaways

This case study highlights two key takeaways for brands looking to appear in LLMs:

1. Prioritize high-buying-intent terms your target audience is likely to use in LLMs. This increases the chances your content is surfaced and cited when generating responses.
   
2. When writing articles targeting high-buying-intent keywords, be extremely detailed about your value proposition, differentiators, and why someone should choose you. This gives LLMs the context they need to present your solution clearly and persuasively.

Improve Your AI Search Visibility with Grow and Convert

Book a call with us to learn more about how we can help your brand show up for high-buying-intent queries in LLMs.

We started working with Level AI in 2024 with the goal of ranking for high buying-intent SEO keywords like “call center customer analytics tools.” In 2025, as AI search exploded in popularity, that goal expanded to include getting LLMs to recommend Level AI when users ask for customer intelligence solutions. 

Across both traditional and AI search, the goal is always the same: drive conversions, not just traffic. That principle guided both Level AI’s SEO and AI search (AEO/GEO) strategy. 

This case study focuses on the AI search component. We’ll show how we earned Level AI brand mentions (not just citations) in 100+ product-centric prompts using our Prioritized GEO strategy. 

If you’d like help improving your AI search visibility, get in touch with us.

How We Achieved AI Search Visibility for Level AI

Prioritized GEO is based on the fact that LLMs almost always search the web to help formulate responses when users ask for product recommendations. This process is known as grounding. Their training data alone is not always up to date or sufficient for these prompts, so they use search to supplement it. 

So, if we want LLMs to recommend Level AI when someone asks for product recommendations in their space, traditional search is the most efficient way to make that happen. For example, if a user asks ChatGPT for voice of customer platforms, it will search the web — and if our article discussing Level AI appears in those results, it significantly increases the chances of being recommended in the response. 

Learn more about Prioritized GEO.

What Keywords We Rank for and Where We Stand in LLMs

Let’s look at a few examples of high buying-intent keywords that are important for Level AI, the content we published to rank for each, where we rank, and the AI prompts we appear in.

Example #1: AI call center monitoring

One of the first keywords we targeted for Level AI was “AI call center monitoring,” and we now rank on the first page for that term.

Our AI visibility tool, Traqer, shows that we consistently appear in the top 5 positions in LLMs when users ask about AI call monitoring solutions. 

Note how Traqer asks about the topic in multiple ways. This is because AI search is not as reproducible as traditional SEO — meaning even slight variations in prompt wording produce different results, and even the same prompt asked at different times can yield different results. 

Traqer accounts for this by asking about a single topic in multiple different ways. We explain this in detail in Topic-Based GEO. 

Side note: We’ve noticed across dozens of brands that visibility on Perplexity and Google AIO often increases before ChatGPT. We hypothesize this is because Perplexity and Google AIO are more heavily influenced by traditional web search results. They function more like search summarizers, so traditional SEO rankings have a greater impact on their output. ChatGPT, on the other hand, blends more of its training data into its responses, which naturally biases it toward larger, more well-known brands. 

Beyond brand mentions of Level AI, we also see LLMs using our article as a source in responses to these prompts.

Here’s an example of Google citing our article and listing Level AI as a solution in its AI Overview for “AI call monitoring solutions.”

Example #2: Call center real time reporting

Another relevant topic for Level AI is “call center real time reporting” because it’s one of the few QA tools that analyzes customer conversations in real time, not after the fact.

Ahrefs shows we rank first for “call center real time reporting” on Google. 

Meanwhile, Traqer shows that we have good AI search visibility across ChatGPT, Perplexity, and Google AIO when someone asks for software suggestions for real-time call center reporting.

We also see that LLMs cite our article more than any other source for topics related to “call center real time reporting.” In fact, across the 6 prompts above, Google AIO, Perplexity, and ChatGPT cite our article a combined 15 times, almost twice as often as any other article. 

This is how you influence LLM responses: create content they discover through web search and use to generate their answers. 

Here’s Google listing Level AI as a popular solution in this space and citing our article first in its AI Overview:

Example #3: Call center analytics software

Another broad topic area where Level AI has a strong value proposition is “call center analytics software.” It’s a popular product category, and appearing in those discussions has long-term value for Level AI.  

So, we wrote an article targeting that keyword, which at the time of writing ranks first on Google.

In Traqer, we see strong visibility for prompts related to call center analytics software on ChatGPT, Perplexity, and Google AIO.

Perplexity recommends Level AI when it’s prompted with questions asking for call center analytics software options, and cites Level AI as one of the main sources it uses to formulate responses.

Example #4: Call center quality assurance tools

The fourth keyword we’d like to cover is “call center quality assurance tools,” where we rank third on Google.

In Traqer, Level AI has strong visibility on ChatGPT, Perplexity, and Google AIO when users ask about call center QA tools.

Although we’re not the top-cited article, LLMs consistently reference us across multiple prompts related to that topic.

We can’t cover every keyword we’ve targeted for Level AI, but these 4 examples demonstrate a clear pattern: ranking well for a keyword on Google gives us AI search visibility for prompts related to that keyword. Keywords we aren’t yet ranking for have little to no visibility in LLMs.

All in all, we’ve targeted 50+ keywords for Level AI so far, resulting in visibility in LLMs for 100+ related high buying-intent prompts.

Examples of AI Using Our Content to Respond to Prompts

Another core principle of our AI search visibility strategy is that the level of detail in your public content helps dictate how LLMs talk about your products and services (full article). 

This means LLMs act like your first salesperson — summarizing your product’s differentiators, value propositions, and arguments, and using them to educate users on what makes you different from competitors and why they should choose you.

As a result, brands need to produce thorough, detailed content that articulates their solution in extreme detail: who it’s for, use cases, features and benefits, pain points it solves, and differentiators versus competing solutions. This way, you feed LLMs your exact value propositions and differentiators so they can make a compelling case for your solution. 

If you publish generic content that echoes what others are already saying, even ranking in LLMs won’t help because it will describe you the same way it describes everyone else, making readers far less likely to convert.

The differentiators and value propositions we focused on for Level AI were:

• Level AI uses semantic intelligence to understand the full context of conversations, making its results more accurate than many alternatives. Most software simply looks for keywords.
• Level AI can detect emotions in customer conversations. Most call center QA software struggles with this.
• Level AI can auto-score all conversations based on the client’s QA rubrics. Teams don’t have to listen to hours of customer conversations and score agents manually.
• Level AI’s deep analysis of conversations uncovers hidden issues and revenue opportunities without surveys, sampling bias, or manual analysis.

As a result, we’d like to share screenshots of LLMs using almost the exact same text in their responses as we use in our articles.

When we ask ChatGPT for accurate AQM software, it emphasizes that Level AI doesn’t just look for keywords but can understand the full context of customer interactions. 

This is known as semantic understanding.

Semantic understanding is the main value proposition we focus on in our articles.

Here, we can see how ChatGPT says Level AI can detect, label, and filter specific emotions like anger, disappointment, worry, and happiness:

We use almost the exact same phrase in our articles:

These are just a few examples, but generally, when ChatGPT recommends Level AI for customer intelligence solutions, it mirrors the same language, value propositions, and differentiators we use in our content.

What This Means for Your Brand: 2 Key Takeaways

1. Target high buying-intent keywords and prompts your target audience is likely to enter into LLMs. LLMs search the web when users ask for product recommendations, so if you rank well for a keyword in Google search, they’re more likely to recommend your brand when users enter prompts related to that keyword.

2. Write highly specific, differentiated content. This gives LLMs the context to position your product or service clearly. Avoid publishing generic content, or your LLM pitch for your brand will sound the same as everyone else’s. Learn more about writing specific, high-quality content.

How to Take This Strategy Further in Your Own Organization

Ranking for high buying-intent keywords with highly differentiated content should give you a strong foundation for LLM search visibility and SEO rankings.

The next step in our Prioritized GEO strategy is Tier 2: citation outreach. This involves identifying the sources AI systems reference most for a given prompt or topic (which you can do in Traqer) and reaching out to those websites to request a mention of your brand. Sites like G2 or Capterra generally require payment, while competitors are often willing to trade mentions.

This works through the same underlying logic as Tier 1: owned content. When LLMs search the web to respond to user prompts, they see your brand mentioned across multiple sources (not just your own website), increasing the likelihood that they mention you more prominently.

That said, there are two mistakes to avoid when doing citation outreach:

1. ​Asking any website in your industry for mentions: We recommend being strategic and focusing on earning mentions in articles that LLMs are already citing.

2. Not controlling the narrative on other websites: We strongly advise writing the product snippet yourself and being specific about your value propositions and differentiators. This way, your messaging is consistent across sources, giving LLMs the information needed to position your solution in a clear and differentiated way.

Increase Your Brand’s LLM Visibility with Grow and Convert

If you’d like us to improve your brand’s LLM visibility and drive conversions, get in touch for a quick call.

In the marketing community, there’s a growing belief that if you rank for these fan-out queries, you’ll consistently get cited by LLMs.

But we wanted to verify and quantify this: How much overlap is there between fan-out query rankings and the sources ChatGPT actually cites? Do the sources cited in a ChatGPT response rank on Google for the fan-out queries? If so, how highly do they rank?

We focused on ChatGPT because, across dozens of clients, it drives the most referred traffic. 

We looked at 100 different buying-intent prompts, each generating 2-4 fan-out queries, and compared the sources ChatGPT cited against the Google and Bing search results for each fan-out query to see how much overlap there actually was.

In total, only 27% of cited sources ranked on Google for the fan-out queries used by ChatGPT. Bing was slightly lower at 23%. Accounting for the ~10% overlap between Google and Bing results, only about 40% of cited sources appeared somewhere in the first 10 pages of either search engine for those queries. 

That means, on average, more than half (60%) of cited sources in ChatGPT answers don’t rank in the first 10 pages of Google or Bing for the fan-out queries.

We also observed a wide range: for some prompts, there was 100% overlap between citations and fan-out query rankings. For others, none of the cited sources ranked within the first 10 pages for those queries.

This data doesn’t invalidate the connection between traditional search rankings and AI search visibility. According to our study, ChatGPT still searches the web for 80%+ of product-related prompts. And as we’ve written about before, these bottom-of-funnel prompts are where AI visibility matters most. 

But it adds to a growing truth about AI search: LLMs are fundamentally less predictable and less “hackable” than traditional SEO.

More specifically, trying to produce single pieces of content to “rank” for specific AI search prompts, in this case by targeting fan-out queries, doesn’t make sense for AI search. It’s fuzzier and less direct than that.

As a result, brands are better off doing authentic marketing — including publishing extremely detailed and thorough content about their products, use cases, features, and benefits — rather than trying to “hack” their way into ChatGPT by targeting specific fan out queries, or as we’ve talked about in previous articles, with tactics like adding FAQ pages, bullet point summaries, or llms.txt files.

You don’t know the true prompts your users are entering. And even if you did, this study shows that targeting fan-out queries gets you cited less than half of the time. So instead of a content strategy aimed at specific prompts, focusing on product-centric content that covers relevant topic areas makes more sense.

Below, we’ll explore these implications further, along with the obvious next question: if only 40% of citations come from the fan-out query SERPs, where does ChatGPT get the remaining 60%?

A Note on Methodology

If you’ve been following us, you’ll know that we believe bottom-of-funnel, product-centric prompts are the only ones worth worrying about for AI search. Top-of-funnel queries are typically answered by LLMs without mentioning or linking to brands or websites. 

All 100 prompts we tested had buying intent (e.g., “best help desk software for small teams,” “top accounts payable automation tools,” or “best standing desk for home office”). The prompts covered B2B, B2C, software, services, and physical products. 

We used the browser UI, not the API, so the data would reflect what a real user sees. 

Of the 100 prompts, 83 triggered a web search, meaning 17% of the time ChatGPT answered without searching the web at all. When a web search was triggered, each prompt generated 2–4 fan-out queries.

To pull Google’s SERPs for the fan-out queries, we used SerpAPI. For Bing, we ran the searches manually because the API returned unreliable results, so the dataset for Bing is smaller.

Finally, we compared ChatGPT’s cited sources against Google and Bing SERPs at the prompt level, by category, and by domain match (vs exact URL match). 

We also analyzed the sources ChatGPT cited that didn’t appear in Google or Bing SERPs to see what we could learn about where those citations come from — and your chances of being one of them.

The Overlap Between ChatGPT Sources and Google’s SERP Varies Wildly from Prompt to Prompt

20 of the prompts had zero cited sources found in the fan-out SERPs — despite the fact that ChatGPT chose to search the web. Only one prompt out of the 100 we tested had 100% overlap. 

A logical follow up question is, are we seeing this wide range of overlap between cited sources and the SERP because ChatGPT relies more heavily on search for some types of prompts vs others (where its training data is enough)? 

So we also grouped them by business type:

While there is some variability, there’s less variation across categories than you might expect. Physical products had the lowest overlap, which is what we’d expect since ChatGPT is likely tapping into features like Google Shopping that weren’t considered in this study.

B2C SaaS also has a weaker overlap between cited sources and the SERP, but, again, this is to be expected because these types of products are typically pulled from app stores rather than their website directly.

This tells us that the variability comes more from how it weighs search results vs training data rather than from what the user is looking to buy. 

We Went 10 Pages Deep, But ChatGPT Doesn’t for Most of Its Sources

Of the sources in our study that did appear in Google’s results for the fan-out queries, page 1 alone accounted for a third of all matches. The top three pages of the SERP contained ~66% of the matches, with a significant decline after that.

This shows what SEOs (including us) have observed and frankly, wished for, which is that ranking higher improves your chances of getting cited.   

But the next important question is if ChatGPT is only pulling from the live web about 40% of the time on average, where is it getting the rest of the sources?

Where Does ChatGPT Get the Other 60% of Sources It Cites?

At the end of the day, where ChatGPT is pulling its sources from is a black box. But we do have a few theories based on what we observed in the data.

ChatGPT May be Running Additional Queries Behind the Scenes

We can only see the fan-out queries that run in our browser. It’s likely, however,  that ChatGPT is running additional searches on OpenAI’s servers that we never see. 

As QueryBurst has documented, it would make sense for OpenAI to distribute searches across multiple servers simultaneously — using different search engines, running different query variations, etc. If it did all of this sequentially in the browser, responses would be painfully slow.

Response Caching

ChatGPT may also be remembering or caching results from past users’ queries to save resources. OpenAI offers developers discounts if they used cache responses in their builds, so it makes sense that they would be using this approach too. If millions of users are asking similar buying-intent questions, it would be cheaper and faster to cache frequently retrieved sources rather than running fresh web searches every time.

Training Data

There’s evidence that ChatGPT retains specific details and facts verbatim from its training data. Research from NeurIPS has demonstrated that large language models can memorize and reproduce training data verbatim. If ChatGPT can remember exact text snippets, it’s conceivable it could also remember exact URLs.

In fact, we found several cases in our data where ChatGPT cited a real, live URL but gave it an outdated title — one that no longer matches the current page. For example:

In another example, ChatGPT cited the same exact URL twice but with different titles, in the same source list. 

Here are the two titles ChatGPT showed: 

The best video conferencing software in 2026 | Zapier

The best video conferencing software in 2025 | Zapier

And, here’s the title that appeared in Google’s SERP for the fan-out query which is also the actual title of the article as of this writing: 

The best video conferencing software for teams in 2026

One explanation for both cases is that ChatGPT is pulling the URL and its old title from training data, not from a live web search. In the second example, one explanation for why it appeared twice is that  it may have pulled the URL from its training data and from a live web search. 

Caching doesn’t explain this either, since OpenAI’s own documentation shows caches typically last 5–10 minutes, and sometimes up to an hour — not months or years, which is the gap between these outdated titles and the current ones.

Hallucinations

In our dataset, we found roughly 10% of citations led to error pages. Recent academic research puts the number higher. A February 2026 study found hallucination rates above 14% for cited sources, and an earlier 2023 study published in Nature documented similar patterns. These are citations that look real — plausible URLs with plausible titles — but point to pages that don’t exist.

So where does all of this leave us? 

GEO ≠ SEO…

The core takeaway is that you can’t use the same mindset or strategies for GEO or AIO that you do in traditional SEO. In SEO, we’re used to the repeatable and predictable idea of “you write a relevant, helpful, quality article targeting keyword X, get your on page SEO right, and build backlinks, and rank for X”. But, LLM answers are way too unpredictable for this same approach to be useful. 

SparkToro’s recent research reinforces this. They found that the brand names that LLMs recommend are wildly inconsistent. Different users can enter the same exact prompt and not only get a different list of recommendations, but also get those recommendations in different orders. 

Add to that the fact that in AI tools, unlike traditional Google, users ask essentially the same question in wildly different ways and the LLMs factor in a ton of deeply personal context with every users’ prompt, and you see how GEO is less controllable and predictable than SEO.

On top of that, you get different fan-out queries for the same exact prompt. The LLM won’t always search the web even for the same prompts. If it does search the web, the amount it relies on that web search versus training data can differ wildly. Also, the model’s capabilities change based on whether the user is paying or using a free account. I could go on, but you get the point.  

And, all of that unpredictability only accounts for roughly 40% of the cited sources. 

That is why a content strategy built around trying to rank for specific queries in order to be cited is impractical and ultimately will be ineffective.

… But SEO is Necessary for GEO

While the nebulous nature of LLMs is uncomfortable for most marketers and business owners, there are patterns to how LLMs operate and actions you can take to get mentioned more often by LLMs. 

Be Available for Training Data

OpenAI admits to using web crawlers both for live web search and for training data, which means you have a chance of getting into ChatGPT’s training data just by putting content on your website. (More on what this content should be below. Hint: It’s not just “easy to read content with digestible chunks”.)

Increase Your Topical Authority

In the SparkToro piece we mentioned earlier, they mention that certain brands are mentioned repeatedly across responses.The top 3 most mentioned brands were cited 64% of the time for the same prompt on ChatGPT.

And our own data suggests ChatGPT may be recognizing that certain domains have more topical authority than others.

If we count all sources that had some page on their domain rank for a fan-out query — instead of requiring the exact URL to match — the percentage of cited sources appearing in Google’s SERPs jumps from 27% to about 50%.

This suggests that there’s some correlation between what ChatGPT and Google consider to be an authoritative site. And, ChatGPT may be favoring domains it recognizes as topically authoritative, even when it doesn’t cite the exact page that ranks. Your domain’s broader presence across a topic may matter more than whether one specific page ranks for one specific query.

One of the most effective ways to build that kind of topical authority is through SEO. Not the “optimize your meta tags and add FAQ schema” kind of SEO, but the kind where you’re consistently publishing detailed, substantive content that addresses the real pain points your customers have. The more comprehensively your domain covers a topic area, the more likely ChatGPT is to recognize you as a relevant source when someone asks about that topic, regardless of which specific fan-out query it runs.

78% of Cited Sources Were Typical SEO-Style Content, Whether or Not They Ranked

One of the big questions floating around marketing circles right now in various forms is whether or not content type needs to change. Do LLM’s even cite traditional marketing or SEO style blog posts such as list posts? And, important for this study, if you produce a traditional style SEO post like a list post but don’t rank yet for your target keyword (or related keywords), do you still have a shot at being cited by ChatGPT for typically similar prompts? Our data says, yes.

For example, here’s a post by Calendly that was ranking for the fanout query “tools for scheduling meetings with clients without back and forth emails scheduling tool”.

It looks very much like a typical SEO-style post that we would write for that keyword. And they put themselves first on the list, which we always recommend. 

In fact, 78% of all cited sources for these high buying intent keywords were listicles, product pages, homepages, pricing pages, etc — essentially webpages selling your products/services, i.e., typical marketing or SEO content.

While this makes sense for the 40% of sources that were in the fan-out query SERPs, what may surprise some marketers is that it was also true of the 60% of sources not found in Google or Bing’s SERPs. The number only went down slightly to ~74%.

Said another way, even the majority of content cited by ChatGPT that were not ranking for fan-out queries looked and felt like content you would publish to rank for those queries: list posts, landing pages, etc.

This is aligned with the theories discussed above of where else ChatGPT could get these citations from, if not from the fan-out queries: caching past searches, or finding similar content by doing web searches when it was last trained. 

This reinforces what we’ve been saying: the best way to influence LLM responses is to produce bottom-of-funnel, product centric content aimed to rank for similar queries on traditional google search. 

What we can add to that understanding with this data is that you don’t need to target and rank for fan-out queries that you can’t even truly find. Instead, creating typical SEO content is the best way to be available for training data and build topical authority, which, again, are your best bets for showing up in LLM answers. 

So, focus on building domain-level topical authority over time. What we recommend and do for clients is to identify your strengths and find Google keywords that map to those strengths. Then, write helpful, relevant, detailed content for those keywords. 

That’s an actual content strategy.

How this Looks for a Real Client, Toro TMS

Toro TMS is a good example of how publishing content builds AI search visibility. They’re a relatively young brand  that hadn’t done much content marketing before we started working with them. Once we began producing product-centric content targeting relevant Google keywords, their ChatGPT visibility followed.

They published thorough, detailed content about their product and the problems it solves, targeted at real Google keywords where their product is a genuine answer. The AI visibility came as a byproduct of high-quality content marketing.

Traditional SEO is predictable. You pick a keyword, create content around it, build links, and over time, you rank. You know which keywords have volume. You know which pages rank. You can track exactly where you stand. It’s measurable and repeatable.

AI search is different. There’s uncertainty at every step. 

1. Prompts: You don’t know exactly what prompts users are entering, because users don’t type short, repeatable keywords. They have long conversations unique to them. 
   
2. Context: Even if you do know the prompts users are entering, you or your AI visibility tool isn’t going to get the same answer as a real user because LLMs factor in a massive amount of user history, memory, and context to generate a response that’s personalized to that specific person. (Read more: Invisible Prompts).
   
3. Repeatability: Even the same user, asking the same prompt repeatedly, will get different responses and citations each time.

So what do you do with all this uncertainty? How do you get those LLMs to recommend your brand and your product when users ask about solutions to their problems in your space? 

The answer relies on this key fact: LLMs still run on content. 

How Search Changes from SEO to AI Search

Here’s the key difference between traditional SEO and AI search.

In SEO, everyone searches with short keywords and largely sees the same results. Yes, some personalization is done based on your browsing history and location, but it’s mostly around the edges.

For example, say we both search for “project management software.” We’ll likely get similar results: 

Seems normal, right? 

But what if you’re at an enterprise company switching from Jira and you want something very specific, and I’m a small business owner currently using pen and paper looking for my first real project management tool? Those details absolutely affect what project management software makes sense for you versus me, but Google doesn’t have any idea. So it serves us the same ten blue links. 

The way we used to deal with this in SEO is going after a few longer tail variants when they exist, like “enterprise project management software” or “project management software for small business.” But even that is insufficient. That doesn’t capture the nuance I mentioned above: that you’re already on Jira and want something specific that Jira doesn’t offer, and that I’m looking for my very first project management software. 

Those are just two random examples. There are really an infinite number of possible specific requests and pain points for a customer looking for any solution. Traditional Google search simply didn’t have a good way of handling all this personal context. 

But AI search does. 

In AI search, each user has a unique, personal conversation with the LLM where they share a massive amount of context. The LLM understands their company size, existing tools, and highly specific pain points. This context comes from both the current thread and a large body of prior chat history, enabling the model to recommend products tailored to their exact situation.

Here is Benji asking Claude about the best project management software “for us.” Look at how Claude knows what our company is, how big, and more. Benji didn’t type any of that into that prompt, but it knows. We call this phenomenon Invisible Prompts. 

LLMs aren’t just surfacing articles from high-authority domains.

They understand what a user needs, search the web for relevant solutions, read those results, combine that with what they already know, and recommend the product or service that best fits the user’s specific pain points. 

That’s a fundamentally different dynamic. And it changes the kind of content you need to produce.

If you want LLMs to recommend your product, you need to give them the details they need to recommend you in the right scenarios. Not generic “what is project management” content. 

You need to publish specific content that covers:

• What types of customers your product is best for
• What specific use cases and scenarios it handles
• What pain points it solves and for whom
• The features that matter and when they matter
• The benefits and outcomes customers experience
• How you compare to competitors and what makes you different
• Case studies that prove you’ve delivered results

Introducing Your GEO Topic Map

Your GEO topic map is a collection of content that teaches LLMs everything they need to know about your product so they can recommend you in the right conversations.

Here’s how we build it for clients:

1. We map every product angle that matters. Your categories. Your competitors. Your use cases. Your target personas. The jobs your product does. The pain points it solves. This becomes your topic map: the full universe of conversations where an LLM could potentially recommend you. Most companies aren’t thinking this way, much less producing this content. 

2. We interview the people who know your customers best. We sit down with your sales team, your customer support team, your product team, your founders, all of the customer facing roles inside your organization. We ask them exactly what users are telling them are their pain points and use cases. Then we extract the specific positioning, features, and differentiators that make your product the right choice in each scenario. These details are not something you can get from Google research. AI generated copy will not by default produce this. The details required to produce this kind of content have to come from the people inside your company who talk to customers every day.

3. We produce content for each angle at the depth of a sales conversation, not an intro guide. There’s this long standing culture in content marketing where everything is written at an introductory level.  “Intro guide to this.” “Beginners guide to that.” That came from everyone chasing high volume SEO keywords that were by nature beginner-leaning. 

But beginner level content won’t help you when your customer asks advanced and specific questions to an LLM. It wants to recommend brands that talk at that customer’s level. So your content needs to meet the customer there. Think about the level of depth and detail that happens in a sales conversation: detailed discussion of features, screenshots, specific scenarios where your product applies, use cases, case studies. That’s how we write this content.

4. We target Google keywords if and when it makes sense. If some of the user pain points we uncover in our interviews with your team match a known Google search term, we’ll target those search terms with our content so you get two inbound channels (traditional search and AI search) in one go.

For most businesses, there are plenty of these product recommendation keywords (what we call bottom of funnel queries) on Google. Keywords like: “Best [category] software,” “[competitor] alternatives,” “how to [solve specific problem].” 

5. We produce case studies. LLMs are smart enough to connect the dots between a story of your product helping someone and their user who is asking them about a similar issue. So it’s our experience and belief that case studies help. They naturally have the elements of specificity and pain point/solution pair elaboration that we’re talking about here. 

6. We track your AI visibility at the topic level. To make this process work, you need to measure success the right way. That means tracking visibility at the topic level, not the individual prompt level. Unfortunately most AI visibility tools (like Peec, Profound, Scrunch) are default to measuring success on an individual prompt basis. But we built our tool, Traqer.ai to be consistent with this topic-based property of GEO. That’s how we track improvement in visibility for clients and ourselves: which topics are we visible in and how is that growing.

Case Study: Constitution Lending

Constitution Lending is a private lender competing against companies with 80+ domain authority, thousands of backlinks, and decades of industry presence.

We built their topic map by interviewing their lending experts, the people who talk to borrowers every day, and producing content covering every specific lending scenario their ideal customers face: LLC mortgages, DSCR loans for specific property types, bridge loans for particular situations.

The result: Constitution Lending now appears in AI search results for 50+ bottom-of-funnel prompts across ChatGPT, Perplexity, and Google AI Overviews — competing with and often outranking lenders that have been in the market for decades.

This didn’t happen because of an llms.txt file or restructured headings (these on-site hacks don’t help LLMs connect your brand to users’ pain points, read more why here). It happened because LLMs found detailed, specific content about Constitution Lending’s products and could match it to user queries.

[Read the full case study →]

Case Study: InnovationCast

InnovationCast makes innovation management software for enterprise companies. 

Innovation software has multiple use cases, from crowdsourcing ideas from a large employee base, to validating ideas, to scouting new technologies.

When we started working with them, they had a limited content and SEO presence. But by producing content in their various topic areas, we’ve built not only an SEO presence that generates qualified leads for them, but also AI visibility

For example, here is their topic based coverage for “innovation portfolio management software”:

And here it is for another, slightly different way customers talk about software like theirs: “technology scouting software”:

Next Steps

If you’d like to talk to us about building your brand’s topic-based GEO strategy, you can reach out here.

If you’d like to read more about our GEO experience, observations and research here is further reading that underpins this Topic-Based GEO approach: 

• Ultra-Specific Content: Why top-of-funnel content doesn’t matter for AI search and why ultra-specific content is necessary
• Prioritized GEO: Why on-page hacks like llms.txt don’t work and why content and web search is the foundation of GEO
• Invisible Prompts: Why you can’t see the full context behind users prompts 

If you’d like more articles like this as we write them, you can join our newsletter. 

Here’s an example: company leadership asks marketing, “How are we doing on GEO/AEO/AIO?” Marketing turns to a tool like Peec or Profound, which spits out a single brand-visibility percentage, usually shown in a graph. Marketing reports back to leadership, “We have 26% visibility!” 

If that number is higher than last week, everyone’s happy and thinks marketing is “doing their job.” If it’s lower than before, leadership assumes something is wrong and wants marketing to change what they’re doing.

Below, we argue that this approach doesn’t make sense because it’s fundamentally the wrong way to think about GEO performance. 

A brand’s “AI search visibility” isn’t a brand-level metric. It’s a function of the specific topics or prompts being measured.

Reporting visibility at the brand level—especially as a single percentage—is, at best, unhelpful and, at worst, outright misleading. 

As we explain below, this mindset can lead to several negative outcomes, including: 

• Manipulation of metrics to make GEO performance appear stronger than it actually is
• A focus on “ranking” for prompts that have little to no business value
• Marketing teams being disincentivized to push harder or pursue more ambitious GEO strategies
• Resources wasted on activities that have no real impact on lead generation or revenue

Visibility Happens at the Topic Level, Not the Brand Level

If you remove the AI visibility tools and their dashboards from the picture, the idea of a single “AI visibility score” falls apart. In the real world, no brand has one. 

AI search visibility isn’t like domain authority in SEO, which represents a real, meaningful concept: how much weight Google’s algorithm assigns to a domain when deciding what to rank. In SEO, a site with high domain authority will, on average, outrank a site with low domain authority, all else being equal. 

There’s no real-world equivalent to a single AI search visibility percentage. Brands have visibility—or don’t—within specific topics. And if a brand has strong visibility in one topic, that visibility isn’t correlated, based on any data we have seen to date, with its visibility in another. 

By its very nature, AI search visibility is a topic-specific concept. 

Topic Level Visibility Is Actionable

More practically, we believe viewing visibility at the topic-by-topic level is how marketing teams learn which actions actually improve AI search visibility. When teams rely on a single, overall visibility percentage instead, those details get buried and the signal becomes misleading.

Here’s a real example from one of our clients, Mirascope, which has strong visibility on the topic of “prompt engineering tools”:

By contrast, on the topic of “context engineering platform,” they have far less visibility:

What does averaging these two together tell the Mirascope team? Nothing useful.

Leadership should think about visibility in these two topic areas completely separately. The actions required to improve visibility for “context engineering platform”—publishing more content on the topic, clearly explaining how Mirascope supports context engineering, and conducting citation outreach on relevant articles—will have no impact on visibility for “prompt engineering tools.”

The SEO analogy applies here. If Mirascope ranked #1 for the keyword “prompt engineering tools” and #60 for “context engineering platform,” what use would it be to average the two and report on a blended rank? No one does this in SEO because it makes no sense. 

Yet this is exactly what happens in GEO, largely because popular AI search visibility tools default to showing a single, overall brand visibility percentage in their dashboards. 

Single Visibility Percentages Create the Wrong Incentives

A single AI visibility percentage often becomes misleading (at best) or downright manipulated (at worst).

Consider what happens when a marketing team starts tracking additional prompts—aspirational topics they aren’t yet visible for but want to be. Adding those prompts, by definition, drives the overall visibility percentage down. 

On the surface, it looks like AI search visibility declined. Leadership may even comment on it or assume marketing isn’t doing their job well. But nothing bad actually happened. Visibility across ChatGPT and other LLMs is exactly the same as it was the day before. The only change is that the team is tracking more topics and being more ambitious about its GEO goals.

Here’s a concrete example. On January 13, the visibility score in Peec for this brand was 26.7%: 

On January 14, we added two new prompts that were relevant to the brand but had no visibility. As a result, the visibility score dropped to 20%: 

In reality, the brand lost no visibility. The metric only looks worse because new prompts were added.

This creates terrible incentives. It encourages marketing teams to be less ambitious. Teams are incentivized to avoid tracking new prompts because doing so makes their visibility percentage look worse. 

That’s the exact opposite of good marketing. Teams should be encouraged to identify new, high-value topics they want to rank for—not discouraged from tracking them.

How Our Visibility Tool Traqer Solves This

This is exactly why we built Traqer to be topic-based.

Topic-level visibility sits front and center in our tool, and that’s how we report on GEO visibility to our clients.

This page is what marketing teams can share directly with executives or clients. At a glance, everyone sees visibility by topic and can quickly understand where their brand is performing well and where it isn’t.

In the Mirascope example above, it’s immediately clear that they perform well for “prompt engineering tools” but have work to do for “context engineering platform.” No one has to dig through reports or click elsewhere to find this information. 

Yes, Traqer also includes brand-level metrics at the top. But instead of collapsing everything into a single percentage, those metrics show the total number of topics where visibility is low, medium, or high. This distinction matters. Adding new topics doesn’t artificially reduce performance elsewhere. 

This approach does two important things:

• It aligns incentives correctly. You can add 10, 50, or 100 new topics or prompts without hurting your overall score.
• It keeps teams focused on what’s actionable: which topics already have strong visibility and which ones need work.

The Right Way to Think About GEO Visibility

Regardless of whether you use Traqer, another tool, or track visibility manually in a spreadsheet, our recommendation is to think about GEO visibility at the topic level. That’s the proper way to approach GEO strategy.

Overall brand visibility percentages are, in our opinion, a lazy way to assess GEO performance. They hide the details that should (1) guide your strategy and (2) show your team whether GEO efforts are actually working.

When you track topic-level visibility, you can:

• Identify which topics need immediate attention
• Allocate content creation resources to the right areas
• Measure progress on specific topics and initiatives
• Avoid false signals from percentage fluctuations
• Make data-driven decisions about where to focus your GEO efforts

It’s the difference between knowing you have “32% visibility” and knowing you dominate prompt engineering conversations but need work on context engineering topics.

One tells you nothing. The other tells you exactly where to focus next.

Further Reading

• You can read the full Prioritized GEO strategy we execute for clients.
• You can read more GEO strategy articles.
• You can check out Traqer for proper AI visibility tracking.
• You can reach out about working with us to improve your GEO.

While it’s great that customers are finding us and our clients through AI search, this creates the problem of reproducibility.  How do you get more of these leads if they’re coming from extremely unique and personal conversations with, say, ChatGPT?

You can’t see these “AI rankings” in your AI visibility software. Peec, Profound, or our tool, Traqer are not going to show you the three-page conversation your recent lead had with ChatGPT. No one has that data (besides OpenAI, who obviously hasn’t revealed it). These are private, highly personalized conversations where the AI tool knows an astonishing amount about the user’s pain points and needs. 

Company after company we talk to is saying they’re getting these leads, but they don’t know how. They don’t know exactly what customers are typing into LLMs, yet the leads keep coming.

We’re calling this phenomenon Invisible Prompts because you can’t see what they are. 

But they’re very real, and as I’ll argue below, they’re only going to grow. 

What Are Invisible Prompts? 

It’s important to clarify that by “Invisible Prompts,” we don’t just mean prompts of a slightly different wording that you happen to not be tracking — like “what are some good content agencies?” versus “find me the best content agencies”. 

It’s not as simple as just thinking of those alternative wordings and sticking them into your AI visibility tool, feeling like you have everything covered. 

This is not what we mean. It’s not a wording issue. It’s much deeper.

The issue is that even if a real user types in the exact wording you are tracking in some AI visibility tool, ChatGPT (or Claude or Gemini) will factor in a massive amount of heavily personalized information about your customer that you can’t predict. This extra personal information about the user will completely change what ChatGPT recommends to them. 

We’ve started thinking about this as the difference between the literal prompt and the effective prompt. The literal prompt is what the user literally types into ChatGPT. The effective prompt includes all the additional information ChatGPT knows about them. 

For example, one of our leads could type “find me a good content marketing agency”. What ChatGPT will actually do to generate an answer is take into account everything it knows about that customer — things like what size business they have, their budget, their industry, what they’ve tried in the past, who is on their team already, and more. 

So, although their literal prompt may just be “find me a good content agency”, their effective prompt is more like a 1000-word essay on their business situation that ends with “…and so taking all that into account, what do you think would be a good content marketing agency for me?”

This is fundamentally not trackable. At least not in the “SEO style” that marketers are used to, where prompts are like keywords, each with some search volume. 

What we’re saying is we are moving to a world where the search volume of most people’s prompts is one. 

Your AI tracking tool isn’t surfacing these prompts to you — even ones like Profound that claim to have data from millions of “real” chats. That data is coming from people who have volunteered (knowingly or unknowingly) to have their web browsing tracked. Tools like Profound summarize patterns in that data, and then say “these prompts” have “this many” searches. So, fundamentally, it’s still the old SEO way of thinking: there are keywords and those keywords have certain search volumes. 

What Profound or any other tool doesn’t know is the deep personalization underneath each prompt. 

That personalization is what’s deciding whether or not you are recommended to the user. 

Examples of Invisible Prompts 

Back in early 2025, Benji wrote about how AI search would reward ultra-specific, bottom-of-funnel content. That post was an early version of this argument: that LLMs would factor in all this personal context when making recommendations. 

In that post, he shared this example of one of our current clients who found us on ChatGPT. We asked her to share the chat where it recommended us. Here is her exact prompt and the beginning of ChatGPT’s answer, where it recommends us: 

Her literal prompt was just: “What are the best SEO companies in your opinion who have driven massive results with endless leads for people in our kind of space?”

But you can see from the response that the effective prompt was longer. Look at how ChatGPT knows her space, knows her goals, knows her constraints. She didn’t actually type that in. 

So here’s the issue: If we track this exact prompt in an AI visibility tool, we won’t see ourselves (Grow & Convert) mentioned, even though in real life, we were. 

I actually tested this. I opened an incognito ChatGPT window and typed in her prompt, and even replaced “our kind of space” with “the business loans and funding space” to help it along.

The response is totally different from hers and doesn’t mention us at all, much less as the top option: 

This is the visibility issue we’re talking about. 

The real prompt — with all the context, nuance, and detail shaping whether ChatGPT, Claude, Gemini, or any other AI recommends you — is far longer, more detailed, and more personalized than you will ever see.

It’s invisible. 

Why Invisible Prompts Are Only Going to Grow 

This phenomenon, from what we can tell, is only growing as AI tools become more and more ingrained in everyone’s personal and professional lives and workflows. More people are logging in and staying logged in (Do you know anyone who uses ChatGPT or Claude regularly but in incognito?). 

ChatGPT, Claude, and Gemini are building longer and longer histories of us. People are even using agents now and giving those agents access to their email and calendar. The context windows in these tools are growing (meaning they remember and factor in more information), and the tools themselves have memory features that remember things from all of your chats when answering new questions. 

In other words, ChatGPT knows more about a user who’s been chatting with it for a year than one who just signed up. That history is context. And context changes recommendations.

All of which means: the gap between the literal prompt and the effective prompt is only getting wider.

How to Get More Invisible Prompt Leads

So what do you do to get more AI search leads if the true prompts are invisible? That’s what every company who observed this asks us. 

Our answer is the same thing we’ve been saying since Benji’s AI search post last year: create ultra-specific content that teaches LLMs all the myriad use cases and pain points where your solution is a good fit. You need to feed LLMs the details they need to recommend you in specific scenarios even though you can’t predict the searches a priori. 

For example, we didn’t know our financial sector client above would ask ChatGPT for an SEO agency, much less how. We didn’t create content specifically for her prompt. We’ve just been writing about how most content marketing does an awful job of generating leads, and how we fix that, for 10 years. 

That’s what you need to do. Create content for as many specific use cases and scenarios as possible. Start with the places and ways you shine. Start with actual case studies. Then move to features, pain points you solve, and the specific situations where your product or service is the right fit.

This Is Why Owned Content is Tier 1 in Prioritized GEO

This is also exactly why we have “owned content” as the highest priority tier in our Prioritized GEO strategy. 

Creating a variety of your own content, as opposed to chasing mentions in Reddit, gives you the room to discuss all the nuance and detail of where and when you shine, what pain points you solve, and more. 

You can do this easily on your own site. On Reddit, what are you going to do? Drop your name in one sentence? That’s not enough context for an LLM to connect you to a specific user’s needs.

What Not to Do: Over-Focus on Single Prompt Visibility

It also explains why we de-emphasize Tier 3, the on-site content tweaks. Adding a schema or an LLMs.txt file won’t help you cover these unique scenarios. An FAQ might help at the margins, but what’s far more effective than that is actual detailed content.

It also means that trying to game the system and show up for a single really specific prompt doesn’t make sense, because no one is actually going to ask that exact prompt. If you produce a dedicated article for one narrow prompt and show up for it, it may look good inside AI visibility tracking software, so your client or management team might be happy. But possibly no one is going to have that exact exchange with an LLM, so it’s not likely to help in a real scenario.

How to Measure Progress  

So is measurement totally hopeless? No. This is all still evolving, but we don’t think so. We’re actually leaning heavily into Topic-Based GEO to solve this, and here’s why.

While you can’t know the exact prompts people will use, you can observe, across clients and over time, trends in how much you’re showing up in AI search at the topic level.

We can see that clients with established brands and years of rock-solid SEO presence in their industry have visibility numbers like this:

While clients that are much younger, with newer domains, and without a long history of content and SEO presence, have numbers like this:

That matters. 

And as per Topic-Based GEO, we can see this across topics. For example, our own AI visibility for “B2B content marketing services” versus “B2C content marketing services” looks completely different:

While we’ve had both B2B and B2C clients and our process works on both, we’ve had more B2B clients, have written more about B2B challenges, and published more B2B case studies. So the industry — and by extension, the LLMs — know us better in a B2B context, and that shows in visibility differences.

These topic-level differences are actionable. It means if we want to get more B2C clients, we need to start producing more content about our B2C work. That way, AI tools will have that information and context and be more likely to recommend us when someone who works at or owns a B2C business asks about content agencies.

That’s how you work invisible prompts in your favor. The prompts themselves may be invisible, but you can still monitor your presence when certain topics are broached.

Which takes me to a final analogy before we close.

Invisible Prompts Are Digital Word of Mouth

This is all curiously similar to the oldest and most effective marketing channel of all time: word of mouth. 

Word of mouth is the holy grail of marketing and brand building. You eventually want to just be recommended whenever someone asks for a product or service in your space. 

But think about what actually happens when you ask your friend for product recommendations. That conversation has all the nuance and detail about your specific situation behind it. 

For example, if I ask Benji for surfboard recommendations, my literal question to him may just be “Hey what board should I buy?” but in his head, he’s going to apply the same complexity that ChatGPT does. He’ll factor in all he knows about me and my surfing: the fact that I’m a beginner, that I’m 6’6” tall, that I currently ride a 10’ foam board, and that I’m extremely handsome. All of it matters. 

So while my literal question to him is short (“What board should I buy?”), the effective question he’s answering is much longer. It includes all the details he knows about my surfing. 

That’s true if someone asks a friend about buying software, hiring a law firm, buying clothes or shoes, or even finding a good coffee shop. 

AI search is similar. Just replace the friend or colleague who knows a lot about you with ChatGPT, Claude, or Gemini. 

The only difference: Benji knows about surfboards from years of actually surfing. ChatGPT only knows what it has read. So if you want LLMs to recommend to you the way a knowledgeable friend would, you need to have published the content that gives them the context to do it.

You can’t predict the invisible prompts. But you can build the content foundation that gives LLMs enough detail to recommend you when those conversations happen.

You see the same AI-generated LinkedIn posts with the same structure and the same lazy arguments. AI-generated comments that add nothing. AI slop blog posts that say what every other blog says, just reworded slightly. It feels like the entire internet is getting worse.

But I think the reason people have such a negative view of AI is because of how most people are using it. They’re using it to try to think less and get out of putting in hard work.

That’s not the best use of AI.

Over the past few weeks, I’ve gone down the AI rabbit hole and I’ve gone from an “AI skeptic” to thinking it’s going to create a renaissance in marketing.

From what I’ve seen, AI is most powerful when you already have expertise, you already know what you want the outcome to be, and you use it as a tool to get there faster or build something you couldn’t have built before. 

The marketers who will win with AI aren’t the ones trying to remove themselves from the process. They’re the ones asking:

What can I create now that I couldn’t before?

And this is where it gets interesting for marketers.

AI Accelerates Whatever You Already Are

Here’s a key thing I don’t see enough people talking about: AI accelerates what you do well and what you do poorly.

If you’re an agency producing bad content with no real strategy, AI will happily help you produce bad content at scale. Congratulations, you can now publish 10x more mediocre blog posts per month.

But if you already have a great process, smart people, and creative ideas, AI unlocks things you simply couldn’t do before. It doesn’t replace thinking. It removes the bottlenecks that used to prevent your thinking from coming to life.

Here’s an example:

I recently wrote a case study we’ll be publishing on Grow and Convert. We like to make our case studies data-driven, not just “traffic went up,” but digging into the data to uncover patterns that actually make the story compelling.

For this particular case study, we had spreadsheets full of conversion data broken out by month and by individual article. In the past, I would have needed someone on my team to help me analyze all of that data, find the interesting patterns, and build the charts and data visualizations. That’s at least two or three people involved: an analyst, a designer, and me writing.

Instead, I did the entire thing myself using AI. 

I had it analyze the conversion data and surface patterns I wouldn’t have spotted manually. I built data visualizations that previously would have required our designer. I queried Ahrefs to compare the keywords the client ranked for before we started working with them versus after, and used AI to help analyze the shift in keyword strategy. 

The result was clear: AI identified that every single keyword they ranked for previously was informational and top-of-funnel, while the keywords we helped them rank for were almost entirely buying-intent. Pairing that with conversion data to quantify the business impact made the case study dramatically stronger.

Had I done this without AI, the analysis wouldn’t have been as thorough, the arguments wouldn’t have been as sharp, and it would have taken me significantly longer to complete.

I didn’t ask AI to “write me a case study about X client” (which is how most people are using it right now). It didn’t know what story to tell or what arguments would resonate. I did. AI just removed the resource constraints that would have made that level of depth impractical.

We’re Entering the Age of the Creative Marketer

For as long as I can remember, creative marketers have been severely undervalued compared to performance and data-driven marketers.

They’d have great ideas but sometimes had challenges bringing them to life. Because to execute anything interesting, you needed other stakeholders involved. A designer, a developer, or multiple people on the team. The ideas were sometimes costly and resource intensive, and the great ideas would die in a Google Doc because there was no way to build them.

AI changes that. If you can dream it, you can build it.

And I think that’s going to unlock a new era of marketing, almost like a marketing renaissance, where the most creative ideas drive the most attention. Not the biggest budgets. Not the largest teams. The best ideas.

The marketers to win in this era will be the ones creating the coolest, most original stuff. Not the ones publishing the most AI-generated volume.

Constraints Have Shifted From “Can I Build This?” to “What Can I Build?”

We used to be constrained by resources. We had more ideas than people or time to execute them.

With AI, that constraint has shifted. We’re no longer limited by execution capacity. We’re limited by the quality of our ideas and our creativity.

That’s a fundamental change, and most people haven’t caught up to it yet.

If all you bring to the table is execution speed — writing fast, designing fast, shipping fast — AI is going to be a problem for you. Because it can do that too.

But when you contribute original thinking, creative ideas, a deep understanding of your customer, and a clear point of view, AI makes you 10x more dangerous. You can bring those ideas to life without waiting on anyone else.

The age of the creative marketer is here. 

The marketers who build the coolest stuff win.

But from what we see, there’s no hard data proving these tactics actually work.

Based on data from 20+ clients, we know that exposure in LLMs comes from our Prioritized GEO strategy: ranking for traditional SEO bottom-of-the-funnel keywords with highly differentiated content.

So, we wrote this article to show an example of how we used Prioritized GEO to help Constitution Lending, a private money lender, appear in the top three AI recommendations for 50+ bottom-of-the-funnel prompts, despite having a domain authority under 40.

How LLMs Recommend Products and Services: They Search the Web When Users Ask for Recommendations

Our AI search strategy is based on a simple reality: all major LLMs search the web to help formulate responses. This process is known as grounding.

In the example below, you can see how Google AIO — the largest AI search tool today by traffic — uses our content to generate its AI summary for the query “LLC mortgage lenders.” This is an important keyword for Constitution Lending, as a large portion of their target market consists of borrowers looking to secure mortgages under an LLC structure.

In this case, Google AIO is literally an SEO summarizer (an “overview” of the search results below), so it’s influenced by traditional SEO rankings. 

Even OpenAI says that ChatGPT searches the web whenever it determines users can benefit from web information — not only when someone explicitly clicks the “search” button.

You can see this when asking ChatGPT for product recommendations. It will almost always say, “Searching the web” (as shown below).

Sometimes, it even tells you what it’s searching for: 

Perplexity also lists its sources when formulating a response, which typically correlate with articles ranking on the first page of Google.

This means AI search visibility is fundamentally linked to traditional SEO. If we rank for “best private lender,” for example, LLMs are more likely to surface our content when searching the web to provide recommendations for private lenders.

In fact, there’s a strong correlation between the keywords we rank for and our visibility in AI search results.

For example, one of our most valuable keywords is “best LLC mortgage lenders,” where we currently rank #1 on Google. We also rank for close variations, including “mortgage loans for LLC” and “mortgage lenders for LLC.”

Using Traqer, the tool we developed to track AI search visibility, we can see that Constitution Lending appears as a top recommendation in ChatGPT (Column #1), Perplexity (Column #2), and Google AIO (Column #3) when users ask for LLC mortgage lenders.

Side note: Nobody knows exactly which prompt users are entering — especially in ChatGPT — because users often have long, multi-turn conversations. As a result, tracking visibility for a single prompt isn’t very useful. Instead, Traqer is built to monitor visibility across topics, where multiple prompts approach the same topic from different angles or with different wording. 

We don’t know exactly how users will phrase a given question, but we can run a range of natural-sounding prompts and measure overall visibility for that topic. If visibility is high, we know we’ll show up regardless of how users ask. If it’s low, we know there’s work to do.

For Constitution Lending, examples include topics like “private lenders” or “LLC mortgage lenders.” We’re also ranking on the first page for “best DSCR lenders,” “top DSCR lenders,” “DSCR lenders,” and “best DSCR providers,” among other variations.

In Traqer, we can see that we’re consistently in the top 1–3 brands cited for prompts related to DSCR lenders.

Here are the exact prompts we’re tracking for that topic and where Constitution is being mentioned or cited as a source:

What’s also interesting is that we see very little AI visibility for keywords we aren’t yet ranking for. 

For example, Constitution Lending recently started offering cash for houses, and we began publishing content on that topic not long ago. Some of those articles aren’t ranking yet. When we check Traqer for “sell my house fast Connecticut,” Constitution Lending rarely appears as a recommendation, showing how a lack of SEO rankings directly correlates with a lack of AI visibility.

These are just three examples, but the pattern is consistent across different keywords and clients. When we rank well in traditional SEO for a keyword, our AI search visibility for related prompts increases. When we don’t rank well in SEO, AI search visibility is limited. 

Below, we walk through our process for ranking in traditional Google search and, as a result, gaining exposure to LLMs.

Our 4-Step AI Search Visibility Strategy for Constitution Lending

Step 1: Identify Bottom-of-the-Funnel, High Intent Keywords

At the start of the engagement, we used Pain Point SEO to identify bottom-of-the-funnel keywords that someone urgently looking for a private lender would search for on Google.

We’ve written extensively about why you should target bottom-of-the-funnel keywords in your SEO strategy. In short, bottom-of-the-funnel keywords convert 10x to 25x better than top-of-the-funnel keywords because the reader is actively looking for a solution and there’s significant buying intent. 

In Constitution Lending’s case, that means borrowing intent (and yes, these keywords are still generating traffic and leads through traditional SEO, even in today’s post-ChatGPT and post-AIO world).

For Constitution Lending, these bottom-of-the-funnel keywords included:

• Best DSCR lenders in Connecticut
• Best LLC mortgage lenders
• Best fix and flip lenders
• Fastest closing private lenders

Articles that rank for these keywords — and therefore gain AI visibility — are far more likely to convert than top-of-the-funnel keywords (e.g., what is a DSCR lender, how do I calculate DSCR), where readers still need to do research before applying for a loan.

Focusing on bottom-of-the-funnel keywords is even more important in the context of AI search visibility because these are the only topics where LLMs mention brands, products, or services in their responses. If you ask an LLM a top-of-the-funnel question like “What is a DSCR loan,” it’ll simply explain the concept without recommending or linking to any brand, because there’s no buying intent.

The old content marketing convention — where a company like Constitution Lending would write content like “What is a DSCR loan” in hopes of getting traffic — is being upended by AI search. Content like that never converted well into leads (as we’ve shown for years), but it at least used to generate traffic. Now it doesn’t even do that. 

Step 2: Prioritize Lower-Competition Keywords Using Underdog SEO

Normally, we’d start producing differentiated content to rank for the bottom-of-the-funnel keywords identified in Step 1.

However, many of the lenders we were competing against had 80+ domain authority, thousands of backlinks, and decades in the lending space. Constitution Lending, by contrast, was a relatively new website with lower domain authority. We knew we were unlikely to outrank these high-authority competitors in the short term by targeting the same bottom-of-the-funnel keywords.

Instead, we applied our Underdog SEO framework to start ranking in Google and AI search results sooner.

Underdog SEO focuses on prioritizing keywords tied to specific use cases or verticals. These long-tail, vertical-specific keywords are typically easier to rank for because larger lenders aren’t targeting them, which means there’s less competition and less tailored content on the web. Since traditional SEO is the path to AI visibility, targeting lower-competition keywords helps us get on LLMs’ radars faster.

For example:

• What we didn’t target first: Best hard money lenders

• What we targeted: Best hard money lenders for first-time investors

• Why? Only one article on the first page is specifically tailored to first-time investors. The rest cover hard money lenders broadly and simply happen to rank for “best hard money lenders for first-time investors” due to limited competition. By creating content specifically tailored to first-time investors, we’re far more likely to rank well.

These are some of the vertical-specific keywords we ended up prioritizing for Constitution Lending:

• DSCR lenders for multifamily homes
• Private money lenders for residential real estate
• Hard money lenders for first-time investors
• DSCR loans for construction
• Commercial loans for rental properties
• DSCR loans for wholesalers and brokers

Step 3: Write and Publish Highly Differentiated Content on Constitution Lending’s Website

With a list of bottom-of-the-funnel keywords prioritized using Underdog SEO, we began drafting long-form blog posts to rank for them.

A common mistake at this stage is hiring freelance writers with little to no domain expertise. This often results in what we call a Google Research Paper: generic content that simply regurgitates what’s already been said online.

This is problematic because AI effectively acts like a salesperson. When LLMs search the web, they consume your content and use it to communicate your value props and differentiators to users. In other words, the content you publish directly shapes how LLMs describe and position your solution. 

For example, below you can see ChatGPT using nearly identical language to what we published in our article:

If your content lacks detail on how you address customer pain points better than the competition, AI won’t have enough information to make a compelling case for your solution. Instead, it will default to the same generic language it uses for every other option on the market.

Additionally, AI develops a deep understanding of users (their context, intent, and pain points) through both current and past conversations. When it searches the web, it looks for solutions that best align with that context. 

If your content doesn’t clearly explain how your solution addresses those needs, LLMs won’t connect your offering to users’ unique circumstances, which ultimately limits visibility.

This means your content needs to clearly articulate your differentiators and speak directly to the pain points your target customers are likely to express to LLMs.

• How are we better than the competition?
• What problems do we solve?
• Who do we serve best?
• In which scenarios or use cases do we truly shine?

With this in mind, we interviewed experts at Constitution Lending who work with borrowers every day and deeply understand their challenges. Those conversations surfaced a few consistent themes:

• Most lenders are slow. Borrowers want fast funding.
• Many lenders reject applications at the last minute. Borrowers want early certainty that they’ll get funded.
• Many lenders require large down payments. Borrowers want lower down payments.

We then created content that directly explains how Constitution Lending solves these problems.

As a result, when a borrower tells an LLM they’re looking for a fast, low-down-payment lender that won’t reject applications at the last minute, LLMs are far more likely to recommend Constitution Lending. We’ve essentially given LLMs the exact information they need to make that connection.

Step 4: Get Brand Mentions on Other Sites That Appear as Citations

Another important component of our Prioritized GEO strategy is securing brand mentions on sites that already appear as citations for the AI prompts we want to rank for. This places Constitution Lending within the articles LLMs actively reference when formulating responses. Brands shouldn’t build citations from just any website in their industry — only from sites that LLMs actually cite.

This works through the same mechanism described earlier. When LLMs search the web for solutions that match a user’s query, appearing across multiple cited sources (including your own site) increases overall AI visibility.

That said, we haven’t begun citation outreach for Constitution Lending yet. So far, we’ve focused on publishing high-quality content on their own site, which has already driven strong AI visibility. In the near future, we may begin outreach to cited websites to secure inclusion in their articles.

Note: We suspect that the lack of brand mentions across multiple third-party sources explains why they appear less frequently in ChatGPT compared to Google AIO and Perplexity. Our data suggests ChatGPT prioritizes diverse brand mentions and places less emphasis on owned content alone. We’ll continue monitoring ChatGPT visibility as we begin citation outreach.

Results: We Appear in the Top 3 AI Recommendations for 50+ Bottom-of-the-Funnel Prompts

Without getting too deep into the specifics of Constitution Lending’s business, our content strategy focuses on four core areas:

1. Lending: Constitution Lending provides financing to borrowers purchasing investment properties.

2. Investing: Constitution Lending allows investors to purchase shares in loans in exchange for borrower interest payments.
   
3. Cash for houses: Constitution Lending buys houses directly, helping homeowners sell in as little as 24–48 hours.
   
4. Non-performing loans: Constitution Lending operates a real estate fund that invests in non-performing loans.

Lending Results

On the lending side of the business, the goal was to get LLMs to recommend Constitution Lending to users when they search for any type of real estate investment lender. 

In the screenshots below, you can see that they’re first in Google’s AI overview for numerous keywords:

If we look inside Traqer, you’ll see that they have strong visibility for several bottom-of-the-funnel prompts in ChatGPT, Perplexity, and Google AIO:

Investor Results

We also create content targeting keywords investors would search for when evaluating real estate loans. 

As a result, Constitution Lending appears as a top recommendation for bottom-of-the-funnel AI prompts such as:

We Buy Houses Results

Constitution Lending recently expanded into the houses-for-cash business, helping homeowners sell quickly at a fair price. They can purchase houses in as little as 24 hours, compared to the months it often takes with a traditional real estate agent.

Our objective, therefore, is to appear in AI search results whenever someone asks for help selling their house fast.

We only began publishing “houses for cash” content a few months ago. Even so, Constitution Lending is already being recommended for several bottom-of-the-funnel AI prompts such as “sell my house fast Waterbury” and “sell my house fast Bristol.”

Non-Performing Loan Fund Results

The non-performing loan fund is another part of the business where we only recently began publishing content. Many of the real estate investment funds we’re competing against are massive — firms like Blackstone and Vanguard that have been around for several decades.

As a result, we expected it would take several months to gain traction in AI search results. However, we’re already seeing AI search visibility for prompts where users are actively looking for a fund.

Increase Your AI Search Visibility with Our Prioritized GEO Strategy

Schedule a quick call to learn how we can help your brand increase AI search visibility for high-buying intent keywords.

In fact, we’ve seen AI visibility companies, agencies, and consultants recommend tactics like adding llms.txt files to your site, rewriting headings as questions, creating FAQ sections, including “key takeaways” at the top of articles, running marketing campaigns on Reddit, and implementing schema markup specifically for AI engines. 

The problem is that most of these ideas aren’t actually strategies but simply a collection of random tactics packaged as the new frontier of search optimization, which gives agencies something new to sell and add to their retainers. Meanwhile, marketing teams are scrambling to implement these disparate tactics without any real evidence that they work.

We’ve tested many of these tactics over the past year, and the results have been underwhelming. Most are unproven, and some, like adding an llms.txt file to your site, have actually been shown (through our testing and others’) to make no measurable difference in AI visibility. Yet they keep getting promoted because they’re easy to sell and technically sound like they should work.

Instead, what we’ve found to work is focusing on getting your brand exposed to LLMs rather than tweaking your content to supposedly help LLMs better digest your content. This finding led us to develop a framework called Prioritized GEO that organizes all of the random AI SEO tactics floating around online into a clear hierarchy of what to prioritize.

In this article, we’ll walk through the AI SEO framework we use to rank and drive visibility for our own clients in AI search, the types of content we create, and the process you can follow to improve your visibility in ChatGPT, Perplexity, Google AI Overviews, and other LLMs.

Our AI SEO Strategy to Increase Brand Visibility in LLMs

Based on our research and client work, we’ve developed a framework called Prioritized GEO to help brands understand where to focus their resources and in what order so they can get their brand mentioned and cited in AI search.

We call it the GEO Priorities Pyramid, and it looks like this:

Every AI SEO tactic we’ve seen fits into one of these three tiers:

• Owned content
• Off-site mentions
• On-site tactics

Owned content is content you write and publish on your own site.

Off-site brand mentions are getting online publications like magazines, industry sites, or communities like Reddit or Wikipedia to mention you (“digital PR”).

On-site tactics is what we’re calling everything that involves changing your site or content to help LLMs better understand it, like adding llms.txt, starting articles with “key takeaways”, wording headings as questions, adding schema and more.

Most of the AI SEO advice online focuses on the top of the pyramid (on-site tactics like llms.txt files and FAQ schema), probably because they’re the newest and most “AI-specific.” But in our experience, those tactics have the least impact on AI visibility.

The foundation of your AI SEO strategy should be owned content that ranks in traditional search. That’s why we have that at the bottom of the pyramid. You can then amplify it with off-site brand mentions, the next tier in the pyramid. Finally, on-site tactics are worth experimenting with, but don’t expect them to move the needle if you’re not already visible in the web searches LLMs are performing.

Below we explain the details behind each of these tiers and why brands should prioritize Tier 1 and Tier 2 over on-site tactics.

Why Being Visible in Web Searches that LLMs Perform Is the Foundation of AI SEO

We spent the past year performing several studies to identify what actually drives AI search visibility. We’ve tested various tactics, analyzed 400+ keywords and prompts across 16 of our own clients, studied which sites LLMs actually cite, and even built our own AI visibility tool (Traqer.ai) to help track clients’ AI visibility because existing tools had too many issues.

We’re not claiming to have everything figured out, but we do have some concrete examples, including our agency and our clients who are consistently appearing in ChatGPT, Perplexity, and Google AI Overviews. 

And the most consistent finding across all of our research is a strong correlation between traditional Google rankings and AI mentions. When we analyzed 400+ bottom of the funnel keywords across 16 clients, we found that when our clients rank on Google’s first page for a given keyword, they show up in ChatGPT and Perplexity responses for that same keyword 77% of the time. When they rank in the top 3 positions, that correlation jumps to 82%.

When they rank in the top 3 positions, that correlation jumps to 82%.

This wasn’t what most of the AI SEO conversation online would have you expect. If ranking on Google correlates this strongly with showing up in AI responses, then the foundation of any AI SEO strategy isn’t some new set of LLM-specific tactics, it’s traditional SEO.

The reason for this correlation comes down to how these AI tools actually work. LLMs like ChatGPT and Perplexity (and definitely Google AI mode or AI overviews) rely on web search to generate their responses, particularly when users ask product-related questions. 

ChatGPT’s web search feature uses third-party search engines, including Bing and Google, and Perplexity’s CEO has publicly acknowledged that they rely on third-party web crawlers in addition to their own. When someone asks these tools for product recommendations or solutions to a problem, they search the web, look at what’s ranking, and synthesize that information into their response.

This means that if your content ranks well in traditional search for the keywords that matter to your business, you’re being exposed to the LLMs that power AI search. In most cases, you don’t need to do anything special to get “crawled” by AI other than continue to work on your traditional SEO and rank for the bottom of the funnel keywords.

We also found a correlation between domain authority and AI visibility. Clients with higher domain ratings were more likely to be mentioned by ChatGPT and Perplexity, which makes sense given that domain authority is itself a function of backlinks and the authority signals that help you rank in Google.

The same factors that make you rank well in traditional search appear to influence whether LLMs mention you. 

But there’s more to it than just ranking. What you rank for and how you communicate your product’s value matter more in AI search than they ever did in traditional search. 

In traditional Google search, you rank for a keyword, the user clicks through to your page, and your page does the selling. The user evaluates your content, navigates your site, and decides whether your product is a fit for them.

In AI search, LLMs don’t just list blue links for users to click through and evaluate on their own. They read your content, synthesize it, and use it to actively recommend (or not recommend) your product to users who are asking for solutions. Your content has become your sales pitch, delivered by AI on your behalf.

This has significant implications for what kind of content you should be creating and how you should be writing it, which we’ll cover in the sections below.

Now, let’s go through each tier in detail: 

Tier 1: Build a Foundation of Bottom of the Funnel Content That Ranks

The foundation of any AI SEO strategy is having content on your own website that ranks in Google for the keywords where you want AI visibility. This is Tier 1 because, as we covered above, LLMs search the web when users ask product-related questions, and if the content on your own website ranks for those queries, you get cited or mentioned by LLMs.

But not all content is equally valuable for AI SEO. The type of content that matters most is bottom of the funnel content: content targeting keywords that indicate the searcher is actively looking to buy or evaluate products like yours.

Why? 

Because in AI search, top-of-funnel content is essentially not going to drive any traffic. This is something we’ve been saying for a while now, but it’s worth repeating because so many content strategies are still built around high-volume, top of the funnel keywords with a focus specifically around generating more traffic.

Consider a typical top of the funnel query in our space: “what is content marketing.” In the traditional SEO world, this was a classic keyword that content marketing companies would target. Ahrefs shows 8,000+ searches a month and a keyword difficulty of 85, meaning it’s competitive with tons of big-name sites ranking for it.

But look what happens when you ask ChatGPT that question. It answers directly and doesn’t link to anything. There’s also no reason for the AI to mention any specific company because the user isn’t looking for a product or service. They’re looking for information, and the AI can provide that information itself.

Google is doing the same thing now. Search “what is content marketing” and you’ll see either a dictionary definition or an AI Overview at the top of the page. The first organic result doesn’t even appear above the fold. Both ChatGPT and Google now default to answering top of the funnel questions themselves instead of sending traffic to a blog post.

For top of the funnel queries, Google might still send you some traffic because it still links to other websites in the AI overviews, but ChatGPT, Perplexity, or Claude will likely not send any. And even the Google traffic is declining as AI Overviews become more prominent. We’ve seen this pattern across our clients where impressions are holding steady or growing, but clicks are declining because AI Overviews are answering queries directly.

This is why you should focus on bottom of the funnel queries because the response for these queries are different. 

When someone asks ChatGPT “what’s the best project management software for remote teams” or “Asana alternatives for small businesses,” the AI can’t just answer with a definition. The user is asking for product recommendations, which means the AI needs to recommend products. This is where your brand can get mentioned.

It also means LLMs will nearly always search the web for these queries because they know their training data alone doesn’t have the most up-to-date product information. They need web search to supplement their knowledge. It’s via these web searches that you can influence the LLMs generated response and get your brand mentioned. 

Here is a screenshot we quickly took of ChatGPT saying it’s searching the web for the prompt above, right before it generated its response. This was without clicking the “search” button when typing in this prompt.

(Also note how the answer to this query above shows citations from “FastrackPR” and “Nova Media Group”. This is in contrast to the “What is content marketing” example above where ChatGPT neither searched the web nor had any citations in its response.)

This is why your AI SEO strategy should focus exclusively on showing up for bottom of funnel, product related prompts and topics, not top of funnel informational topics. 

At Grow & Convert, we’ve actually  been focusing on bottom of the funnel content since 2018, long before ChatGPT and AI search were popular, for a separate reason: traditional SEO traffic from bottom of funnel keywords converts far, far higher than top of funnel traffic. We published our original article called Pain Point SEO explaining why. 

We showed through multiple case studies that bottom of the funnel keywords convert 10x to 25x better than top of the funnel keywords. Now, in AI search, this gap is even wider because as we explained above, top of the funnel content often gets zero brand visibility for you (the user gets their question completely answered by the LLM tool so unlike traditional SEO you don’t even get the user to your site to get even the possibility of them converting).

So, we suggest you do the same and focus on creating content to rank for bottom of funnel topics so that your brand and content gets mentioned and cited for related prompts in AI search. 

Types of bottom of the funnel keywords you should be targeting

We cover these extensively in our article on SaaS content strategy, but here’s a summary with examples:

Software category keywords are searches like “best [use case] software,” “[industry] software,” or “[use case] tools.” These are people actively looking for a product in your category.

For example, if you sell time tracking software, keywords like “best employee time tracking software,” “time clock app with GPS,” or “construction time tracking software” would fall into this category. The variations are extensive and you can combine use cases with industries, business sizes, integrations, and more to find dozens of keyword opportunities.

Comparison keywords include “[competitor] alternatives” and “[brand] vs. [competitor]” searches. These are people who are evaluating options and are very close to making a decision.

For example, “Asana alternatives,” “Monday vs. Asana,” or “Hubspot alternatives for small businesses.” If you’re a smaller brand competing against well-known players, comparison content is particularly valuable because you’re inserting yourself into a conversation that’s already happening.

Jobs-to-be-done (JTBD) keywords are searches where someone is trying to accomplish a specific task that your product can help with. These often start with “how to” but the searcher isn’t necessarily looking for software yet.

For example, “how to plan delivery routes for multiple stops” or “how to track employee hours.” The content that ranks for these keywords typically explains how to do the thing manually and then introduces your product as a better solution.

Use case and template keywords are related to jobs-to-be-done but more specific. Searches like “daily standup meeting template” or “delivery driver training checklist” indicate someone is trying to solve a specific problem.

You can rank for these by providing a free template and then explaining how your product makes the template unnecessary or more effective.

Write Content That LLMs Will Actually Use to Recommend You

How you write content matters as much as what keywords you target. This is where AI search differs most from traditional SEO, and it’s something most marketers haven’t fully internalized yet.

In traditional SEO, your content’s job is to rank and then get clicked. Once someone lands on your page, your website does the selling. The user reads your content, navigates your site, checks out your pricing page, maybe looks at a few case studies, and eventually decides whether to sign up or request a demo.

In AI search, that whole process gets compressed because when someone asks ChatGPT for product recommendations, the AI reads your content (and your competitors’ content), synthesizes it, and delivers a recommendation in its own wording directly to the user. Look at the screenshots above about project management software. ChatGPT is writing its own bullets selling ClickUp and Asana. So in AI search, your content is not just helping you rank but also teaching the LLMs how to position or message your brand or products.

Effectively, LLMs are acting as your SDRs (sales development reps). They’re reading your content and using it to pitch your product to potential customers. So, if your content doesn’t clearly communicate who your product is for, what problems it solves, why it’s better than alternatives, and what makes it different, AI engines won’t know when or how to recommend you.

This is why specificity is so important in AI search. Generic content that could apply to any product in your category won’t help you. 

You need content that clearly articulates:

• Who your product is for (specific industries, company sizes, use cases, personas)
• What specific problems it solves (not vague benefits, but concrete pain points)
• How it solves those problems differently than competitors
• Proof that it works (case studies, specific results, customer examples)

For example, when someone asked ChatGPT for “best pain point focused SEO agency”, it recommended Grow & Convert and described us as an agency that  “prioritizes content that solves specific buyer problems and targets bottom-of-funnel keywords with clear purchase intent.” 

That description obviously comes from our site. We’ve talked about prioritizing bottom of funnel content for a decade on Grow & Convert and our most popular framework is Pain Point SEO, which is where ChatGPT gets the line “solves specific buyer problems”.. The AI read our content, understood our positioning, and used it to match us with a user whose needs aligned with what we do.

The companies that get recommended most consistently by LLMs are the ones that have published content clearly explaining who they’re best for, what they’re best at, and why they’re different from everyone else.

This is also why AI-generated content fails for AI SEO. It sounds counterintuitive, but content written by AI is actually the worst kind of content for getting recommended by AI.

The reason comes back to how LLMs work. They’re trained on massive amounts of text from the internet, and they’re designed to produce responses that are similar to their training data.

When you use AI to write content, you get output that reflects the most average, most common way of saying things. It’s grammatically correct and sounds articulate, but it doesn’t say anything original or different. So if this is what you’re publishing, then when ChatGPT et al are trying to pitch your brand to users, it’s going to say generic, unoriginal, undifferentiated things about your products or services. 

We wrote about this years ago when we coined the term Mirage Content (content that looks good on the surface but is really just high-level fluff that regurgitates what everyone else has already said). We argued back then that most content written by freelance writers was Mirage Content.

AI-generated content is essentially Mirage Content at scale. It’s fast to produce, but it doesn’t contain any original thinking, unique perspectives, or specific details about your product that would help an LLM understand when to recommend you.

The content that performs best for AI visibility is content written by humans who actually understand the product, informed by interviews with people inside your organization (sales, founders, product leaders) who can speak to specific features, use cases, and differentiators that you won’t find anywhere else online. This is the content that helps LLMs understand what makes you different, and it’s the content that gets recommended.

Tier 2: Build Authority with Off-Site Brand Mentions

The next tier of your AI SEO strategy is getting your brand mentioned on other websites. This is what we call off-site brand mentions, and it includes tactics like digital PR, citation outreach, guest posting, and link building.

Off-site mentions matter for AI visibility for the same reason owned content matters: LLMs search the web to formulate their responses. When someone asks ChatGPT for product recommendations, it doesn’t just look at your website. It looks at what other sites are saying about products in your category and if your brand is mentioned on sites that rank for the keywords you care about, you’re more likely to show up in the AI’s response.

But not all off-site mentions are equally valuable. If you’ve followed the AI SEO conversation online, you’ve probably seen advice to get your brand mentioned on Reddit, Wikipedia, and Forbes because these are supposedly the sites that LLMs cite most often. Infographics showing the “top cited domains by AI” usually accompany this advice.

The problem with this thinking is that it assumes there’s a universal list of most-cited sites that applies to all businesses. There isn’t.

When someone asks ChatGPT “what are the best dispatch software options for trucking companies” or “what are some good alternatives to Asana for project management,” ChatGPT doesn’t pull just look for answers on a static list of the most popular sites on the internet. It searches the web for that specific query and cites whatever sources are most relevant.

Our research on which sites LLMs actually cite found that for product-related queries, LLMs cite industry-specific sites 86% of the time and generic sites like Reddit only 16% of the time.

This makes sense when you think about it. If you’re asking for trucking dispatch software recommendations, the AI is going to cite sites that review trucking software, not a random Reddit thread.

This means your off-site strategy should focus on getting mentioned on the sites that actually rank for the keywords you care about. 

Don’t waste time trying to get mentioned on Reddit or Forbes just because someone told you LLMs cite those domains a lot. Instead, look at the specific prompts and keywords where you want AI visibility, see which sites LLMs are actually citing for those queries, and focus your efforts on getting mentioned there.The process is straightforward.

Take the keywords you identified in Tier 1 and enter them into ChatGPT, Perplexity, Google AI Overview, or any other LLM you’re interested in. Look at which sources get cited in the responses. These are the sites you should be targeting for off-site mentions.

For example, here is such a list from our AI tracking tool, Traqer:

Citation outreach is how we execute this for clients as a part of our engagement. We get a list of the sites that LLMs cite for our clients’ target keywords via Traqer, and we reach out to those sites to try to get our clients mentioned in their content. 

Sometimes this means getting added to an existing listicle or comparison article, contributing a quote or expert comment, or writing a guest post. The specific tactic varies, but the goal is always the same: to get mentioned on sites that rank for keywords where we want AI visibility.

Why Are Off-Site Mentions Tier 2 and not Tier 1?

The key distinction between Tier 1 and Tier 2 is control. With owned content (Tier 1), you have complete control over what you publish, how you position your product, and how detailed you get. With off-site mentions, you’re dependent on other sites to mention you accurately and favorably.

This is why we put owned content on Tier 1 and, therefore, slightly higher priority than off-site mentions. Both your content and mentions in others’ content work via the same mechanism: helping expose your brand to LLMs when they search the web to answer users questions (a process called “grounding”). 

But your owned content is where you have the space and freedom to tell your full story: who you’re for, what problems you solve, how you’re different from competitors, and proof that you deliver results. Off-site mentions help to amplify that story by getting your brand name in front of LLMs through additional sources. 

Tier 3: On-Site AI SEO Tactics

Finally, we get to the top of the pyramid: Tier 3. This is what we think should be your last priority but ironically is where most of the AI SEO advice online is focused: on-site tactics designed to help LLMs better understand and process your content. These include things like adding llms.txt files to your site, structuring headings as questions, adding FAQ sections to pages, including “key takeaways” at the top of articles, and implementing schema markup specifically for AI engines.

We put these tactics at the top of the pyramid (meaning lowest priority) for a reason. They are purported to help AI search visibility in a fundamentally different way than Tiers 1 and 2: the aim of these on-site tactics is to help LLMs better understand or digest the content on your site. 

They don’t help expose your brand to LLMs.  

Also, in our testing over the past year, most of these tactics have made no measurable difference in AI visibility. That doesn’t mean they’re completely useless, but it does mean your time is better spent on getting your own rankings (Tier 1) for queries LLMs may search for when generating responses or getting brand mentions in other sites that are ranking (Tier 2).

For example, llms.txt is a proposed standard from Jeremy Howard, the developer and founder of fast.ai, that would help LLMs crawl and understand a site’s content. The idea is similar to robots.txt but designed specifically for AI crawlers. It sounds logical in theory, but there’s no evidence that OpenAI, Anthropic, or other major LLM providers actually use it. Howard simply proposed it as an idea that he thought would be good for the industry to adopt. Our tests suggest it doesn’t make a difference, and we don’t recommend spending time on it.

Structuring headings as questions is based on the idea that LLMs process question-and-answer formats better than declarative statements. Again, it sounds plausible, but we haven’t seen any data showing it improves AI visibility. If your content naturally lends itself to Q&A formatting, fine. But rewriting all your headings as questions just for AI optimization isn’t worth the effort, in our opinion. Plus the entire magic of LLMs is that they understand natural language as well as humans do, so it makes sense that they don’t need hand-holding in the form of reformatting your content in a specific way.

Adding FAQ sections and key takeaways falls into the same category. The logic is that these formats make it easier for LLMs to extract and summarize information. We’ve tested articles with and without these elements, and we’ve seen clients show up in AI responses consistently with content that has none of these tactics applied. So while they might help marginally, they clearly aren’t necessary.

Schema markup for AI is another tactic being pushed by some agencies. The idea is that structured data helps LLMs understand your content the same way it helps Google understand your content. This may have some validity, but we haven’t seen enough evidence to prioritize it over the tactics in Tier 1 and Tier 2.

Tier 3 tactics are fine to experiment with, but we strongly believe that your focus should be on Tiers 1 and 2. If you have a solid foundation of bottom of the funnel content that ranks and you’re actively building off-site mentions on relevant industry sites, then sure, test some FAQ sections or key takeaways and see if they move the needle. 

Think of it this way: Tier 3 tactics are meant to help LLMs better digest your content. If LLMs aren’t finding your content in the first place (because it doesn’t rank), making it easier to digest won’t help you. The priority is getting your content in front of LLMs, which happens through traditional SEO (Tier 1) and off-site mentions (Tier 2). Only then does it make sense to optimize how that content is structured.

How to Measure AI SEO Performance and Results

Measuring the results of your AI SEO strategy is one of the more challenging aspects of this work, and it’s only getting harder as AI becomes more integrated into how people search and discover products.

Traditional SEO attribution methods are breaking down. For years, the measurement model was straightforward: someone searches a keyword, clicks on your organic result, lands on your page, and eventually converts. You could track this journey through Google Analytics and attribute conversions to specific pages and keywords (via the Google/organic source/medium, for example).

AI search disrupts this model in several ways. When someone asks ChatGPT for product recommendations, they might see your brand mentioned, open a new tab, Google your company name, and land on your homepage. That conversion shows up as branded organic traffic or even direct traffic, not as a conversion from the AI mention that actually drove it. The AI did the work of surfacing your brand, but your analytics won’t reflect that.

Google AI Overviews create similar attribution challenges. When your brand gets mentioned in an AI Overview, users might click through to your site, but they might also just note your brand name and search for it separately later. Or they might click one of the links in the AI Overview, which could be your homepage, a product page, or a blog post, depending on how Google constructs the search results for that query. What used to be a clean path from keyword to landing page is now fragmented across multiple possible user journeys.

We’ve written about this in detail, but the short version is that you’re likely underestimating how much AI is contributing to your conversions if you’re only looking at traditional attribution models.

The most reliable way to measure AI visibility impact is to ask people directly. 

We recommend adding a “how did you hear about us” field to your lead forms and demo request forms. This gives prospects a chance to self-report whether they found you through ChatGPT, Perplexity, Google, a referral, or another channel. We’re also encouraging clients to have their sales teams ask this question during discovery calls.

This isn’t a perfect measurement method. People don’t always remember exactly how they found you, and they might say “Google” when they actually started with ChatGPT and then searched your brand name on Google. But it gives you directional data that traditional analytics can’t provide, and we’ve found that more and more prospects are explicitly mentioning ChatGPT and Perplexity in their responses.

You should also track referral traffic from AI platforms directly. 

ChatGPT, Perplexity, and other AI tools do send some clickable traffic, and you can see this in your analytics. It won’t capture all the AI-influenced conversions (for the attribution reasons we mentioned above), but it gives you a baseline. We’ve built reporting that specifically identifies referral sources from LLM platforms so we can see how much direct traffic is coming from AI tools.

Track AI visibility itself, not just traffic from AI. 

This is also why we built Traqer.ai. We needed a way to track whether our clients are showing up in AI responses for their target keywords, regardless of whether that visibility translates into trackable clicks.

There are other AI visibility tracking tools on the market now, but there’s a problem with how most of them work. A lot of these tools generate their own list of prompts and tell you which ones you’re “ranking” for. But these prompts are often random and made up by the tool. You have no idea if anyone is actually typing them into ChatGPT, or if they’re even relevant to your business. That’s letting the tool lead the strategy instead of the other way around.

The right approach is to decide which topics and keywords matter for your business first, then track visibility for prompts related to those topics. You decide the prompts based on your bottom of the funnel keyword strategy, not some tool generating random queries.

With Traqer, we start by identifying the bottom of the funnel topics that matter for each client, then track visibility across multiple prompts related to those topics in ChatGPT, Perplexity, and Google AI Overviews week over week. This lets us see what’s actually working and adjust accordingly.

Track Overall Site Conversions

You should also start to measure conversions holistically rather than just attributing to individual blog posts. 

In the past, we focused heavily on tracking which specific articles drove conversions. That’s still valuable, but it’s no longer the complete picture. We’re now also tracking organic traffic and conversions to homepages and product pages, because AI visibility often drives users to those pages rather than the blog content that actually influenced the AI’s recommendation.

The goal is to look at the overall trajectory of organic performance (rankings, traffic, and conversions across your whole site) rather than trying to attribute every conversion to a specific piece of content. If your rankings are growing, your AI visibility is growing, and your overall organic conversions are growing, your AI SEO strategy is working, even if the attribution is messy.

The brands that are succeeding with AI SEO right now are the ones that understand this measurement challenge and don’t expect perfect attribution. They track what they can, ask prospects directly, monitor AI visibility over time, and look at the overall trend rather than trying to tie every lead to a specific source.

Want to work with us or learn how to implement this AI SEO strategy?

• Our SEO + GEO Agency Service: You can learn more about working with us here.
• Our Content Marketing Course: Individuals looking to learn how to grow their SaaS business with content can join our private course, taught via case studies, here. We include lots of detail and examples not found on this blog. Our course is also built into a community, so people ask questions, start discussions, and share their work in the lesson pages themselves, and we and other members give feedback. We also get on live Zoom calls about once a month and dissect members’ actual content strategies and brainstorm ideas on how we’d form content strategies for their businesses.

We’re not seeing such extreme declines in organic traffic across our 20+ clients, but yes, there’s been some drop.

So in this post, we want to report on that data and answer two key questions:

• What’s actually causing this decline — AI Overviews, ChatGPT, or something else?
• Should marketing teams be worried? And if so, what should they do? 

The Data: Traffic Is Down but Rankings and Conversions Are Growing

We’re seeing a consistent pattern across several clients: more rankings but less traffic. 

Here’s a typical example from one client’s Ahrefs rankings in October 2023, 2024, and 2025:

And here’s that same data shown in a bar graph, so you can easily see the growth in rankings:

But look at this client’s traffic, measured by clicks in Google Search Console (displayed via SEO Gets). The blue line represents clicks:

There’s a slight decline in clicks from mid-2024 to October 2025. 

Notably, impressions (purple line) are trending up. That means our organic listings on the Google SERP are showing up more often now than before, but earning fewer clicks (blue line). The green line, click-through rate (CTR), confirms this with a sharp drop around March 2025. 

We’ll discuss why that date is important below, but this is the trend we’re seeing across multiple clients. 

For instance, here are two more clients’ Search Console graphs showing impressions (purple) increasing while clicks (blue) decrease: 

Finally, what’s interesting is that despite the drop in traffic, clients are seeing steady growth in conversions. 

Here’s data on free-trial conversions attributed to our content for a SaaS client from early 2025 to now:

What’s going on? 

How are businesses seeing record-high rankings, declining traffic, yet steady (or even growing) conversions?

In all our years running Grow and Convert, rankings, traffic, and conversions have always grown together — when rankings and traffic were down, conversions were also down.

But now we’re seeing two metrics rising (rankings and conversions) while traffic falls.

So, why is that?

Why We Think It’s AI Overviews, Not ChatGPT, Causing the Traffic Decrease 

We believe the decline in clicks and site traffic is due to Google’s AI Overviews having more prominence in search, not from users moving off Google for ChatGPT. 

Why ChatGPT Is Likely Not Causing the Traffic Drop

First, in the Search Console data above, impressions are holding steady — and for many clients, they’re actually growing. If the organic traffic decline was caused by users shifting from Google to ChatGPT, we’d expect to see overall impressions drop too. We don’t.

Second, in our clients’ analytics data, traffic coming from ChatGPT is less than 5% of what they get from Google Organic. 

Here’s that same data as a bar graph, to make the difference visually clear:

Ahrefs created a site called ChatGPT vs. Google, and they’re largely seeing the same thing — traffic from ChatGPT remains minimal compared to Google, Bing, and other traditional search engines.

A study from Semrush also shows that ChatGPT is now used as one step in the search process. People might start their search in ChatGPT, but they still turn to Google to complete their searches. So ChatGPT usage alone doesn’t explain the traffic decline.

You can read the full study here.

So if ChatGPT isn’t the cause, what about Google’s AI Overviews?

Why Google AI Overviews Is Likely Causing the Traffic Drop

Notice in the graphs above there’s a sharp drop in clicks (blue line) and CTR (green line) around March 2025. 

I was curious about that timing, so I asked Google about the rollout of AI overviews, and March 2025 seems to be when they rolled it out more broadly:

This strongly suggests that the drop in clicks (and traffic) starting around March 2025 is related to AI Overviews. 

The next question is how and why AI Overviews can reduce traffic without hurting impressions, even as rankings continue to rise.

Why AI Overviews Cause a Decline in Organic Traffic Even as Rankings and Conversions Grow

First, for top-of-funnel queries you used to rank for, AI overviews now answer the user’s question directly — so they no longer need to click into your article. That’s likely the primary reason brands are reporting drops in organic traffic. 

It also explains why we at Grow & Convert are seeing smaller declines in traffic (10%–20%) compared to others reporting 40%–50%. We focus on ranking clients for bottom-of-funnel search terms, meaning queries where users are looking for a product or service like theirs. For those terms, even if an SEO article doesn’t get as much traffic as before, the brand still gets mentioned in the AI overview itself. 

For example, here’s the AI Overview for the term “best call center quality assurance software,” where our client Level AI is mentioned in the AI Overview, and a page on their site is one of the cited sources on the right:

A user looking for software like this isn’t really incentivized to click on the articles on the right — the AI Overview already summarizes which brands are popular. That’s the simple, obvious explanation for the drop in traffic. But there’s also something more subtle happening here. 

Specifically, notice how Level AI is mentioned multiple times in the AI Overview. That means several different user paths could unfold.

Scenario 1: New Tab

The user might just open a new tab, Google “Level AI,” click into the homepage, and request a demo or convert. That conversion would appear as organic traffic to the homepage — not the article — even though the article was instrumental in getting Level AI featured in the AI Overview in the first place. As we’ll discuss below, that creates a major attribution problem. 

Note that if the user’s eyes go to the second bolded “Level AI” mention under “top call center QA software”, there’s no link, which makes it even more likely they’ll just open a new tab and search “Level AI.” 

Scenario 2: Dotted Line Link

Notice how the first mention of Level AI in the AI Overview introduction has a thin dotted underline. When you click that link, it triggers a “disguised” branded search. 

It looks like a normal branded search, but if you check the URL and the “showing results for” line, you’ll see it actually searches for “Level AI call center quality assurance software.” 

As a result, the Level AI homepage isn’t the top result — instead their Quality Assurance Software page appears first, followed by a related blog post, and then the homepage. And all three of those organic links are preceded by a sponsored spot because, well, Google’s gotta make money. 

When this happens, traffic from that search term can show up as any of the following: 

• Branded paid (if the user clicks any of the sponsored links)
• Organic to the feature page 
• Organic to the blog post
• Organic to the homepage

It’s confusing! 

What this means is that traffic that used to be easy to attribute (one search term corresponding to one URL) is now spread across a mishmash of possibilities: users might go to the homepage, a feature page, a paid ad, or a blog post. 

Which brings us to the real consequence for marketers: attribution is getting even harder. 

Attribution Needs to Change Because You’re Not Measuring the Full Benefit of SEO Anymore

SEO attribution has always been tricky, and since the start of our agency, we’ve aimed to fix that. We’ve made a deliberate effort to track leads all the way through the funnel so we can see which blog posts convert best.

We’ve always said our attribution represents a lower-limit estimate of the actual number of conversions our services generate, mainly because there are a number of scenarios where you can’t measure traffic all the way through conversion.

Specifically:

• If a user comes in through SEO but has cookies blocked, your SEO/content efforts won’t be attributed to that lead.
  
• If someone discovers your brand through SEO but switches devices before converting, that conversion won’t be attributed to SEO.
  
• If someone finds your brand through SEO and tells a coworker or friend who later searches for it separately, that lead won’t be attributed to SEO.

Now, with the rise of AI overviews, there’s yet another attribution challenge. If people discover your company through an AI overview, what was previously attributed to organic search now often appears as direct or branded traffic.

Why?

Because, as shown above, AI overviews display company names as clickable links that open branded searches. Many users might also open a new tab and type your domain directly or search for your brand manually. 

In other words, SEO is still driving visibility — it’s what gets your brand mentioned in the AI overview — but the resulting increase in branded and direct conversions isn’t being credited to SEO, even though it originates there.

What’s the Best Way to Account for This Change in SEO?

We’re advising clients to add a “How did you hear about us?” field to their lead forms. This lets prospects self-report whether they came from Google, AI, referrals, or another channel — providing another way to attribute leads. We’re also encouraging sales teams to ask the same question during qualification calls.

In addition to tracking organic traffic to our articles, we’ve started measuring organic homepage traffic as well. We’ve also created additional reports that identify referral sources from LLM platforms to see how much traffic is coming from LLMs.

Overall, we’re resetting expectations around traditional organic leads and traffic metrics and looking at organic leads more holistically — across both homepage and article traffic.