top of page

What AI Says About Your Brand (And Why You Can't Just Track It)

6 hours ago
10 min read

Someone on your leadership team just walked into your office and asked a simple question. "What does ChatGPT say about us?"


Now you feel the pressure. You need an answer. You probably need a number to track. You can't reliably show up in the same spot in AI answers, because they change almost every time. What you can control is what AI learns about you: what your site says, what the rest of the web says, and whether they agree.


The urgency behind your executive's question is real. In a 2024 prediction, Gartner estimated that traditional search engine volume would drop 25 percent by 2026, with search marketing losing market share to AI chatbots and virtual agents. Buyers are changing how they look for answers. Pew Research found that users clicked a traditional result in just 8 percent of visits that included an AI summary, compared with 15 percent of visits without one.


You have likely tried AI visibility tracking tools or run test prompts yourself. It feels like traditional search all over again. You want to show up first. You want to be the definitive answer.


But treating AI like a search engine is a trap. If you optimize for a position that changes every time a buyer asks a question, you will spend your budget chasing ghosts.


AI doesn't rank brands. It learns about them.



Search engines give you a stable page of results. Generative AI gives you a slot machine.

When marketers see a new technology, they immediately try to box it into an old dashboard. We invent new acronyms to make ourselves sound smart. We talk about Generative Engine Optimization or Answer Engine Optimization. As we noted in SEO's Identity Crisis Is Self-Inflicted, the industry loves to rebrand the basics to sell more software.


AI tools do not have a static index of ranked brands. They predict the next logical word based on their training data and the specific context of the user's prompt. In a recent conversation about how Most Decisions Aren't a Tattoo, Stephan Bajaio pointed out a hard truth. AI rankings aren't real ranks, and an answer's second mention can be wrong.

If you ask an AI tool for the best enterprise software for a specific problem, it might list your brand first on Tuesday. On Wednesday, it might list you third. On Thursday, it might leave you out entirely.


You cannot optimize for a rank that does not exist. You can only optimize the information the model consumes. Change the question from "What does AI say about us?" to "What can AI learn about us?" The first question puts you at the mercy of an unstable output. The second gives you levers you can actually pull.


What the research says about AI visibility tracking



Most content about AI brand visibility comes from vendors selling tracking software. They want you to believe you can monitor your way to success. In a Gold Rush, Sell Pans and Picks. But the data tells a different story about how these models actually behave.


Treat all third-party AI statistics as directional, not precise. Here is what the current research shows:


  • AI answers are highly unstable. A November to December 2025 study by SparkToro and Gumshoe tested this extensively. Across 600 volunteers and 2,961 runs, they found a less-than-1-in-100 chance that ChatGPT or Google's AI would return the exact same list of brands across repeated runs of the same prompt.


  • Third-party sources drive the narrative. A March 2025 Pew Research study of 900 US adults found that Reddit, Wikipedia, and YouTube are the most cited sources in AI Overviews. The models do not just read your website. They read what other people say about you.


  • Buyers are researching without you. Long-running Gartner research shows B2B buyers spend just 17 percent of their time meeting with suppliers. If there are 3 potential suppliers, you get about 5 percent of a customer's time. They spend the rest of their time reading the web and asking AI.


If the answers are not consistent, a rank means little. What stays the same is the underlying data the model learns from.


That is why two measures are more useful than a fake "position" number: share of voice and citation rate.

  • Share of voice is how often your brand appears in relevant AI answers compared with competitors for a defined set of prompts.

  • Citation rate is how often your brand, or the sources that mention you, show up in the citations and source links attached to those answers.


Neither metric is perfect. Both still move around. But they are better than pretending there is a stable podium and your job is to stand on it.


Self-audit: what could an AI learn about your company today?


Before you buy a tool to track your brand presence in AI answers, you need to know what raw material the AI has to work with.


Ask yourself these questions. Be honest. If you want to turn this into a formal process later, you can look at gathering Human Proof. For now, just score yourself yes or no.


Level 1: Your site


Level 2: The rest of the web

  • Are you actively mentioned on the review sites, forums, and publisher pages your buyers trust?

  • Do your executives have a footprint on third-party sites, podcasts, or industry publications?

  • When a partner organization lists you on their site, do they describe you accurately?


Level 3: Agreement

  • Does your website's description of your services match what your customers say in their reviews?

  • If an AI reads your site and then reads a Reddit thread about your brand, will it find the same core message?

  • Are your internal experts saying the same things in public that your marketing materials claim?


If you answered no to more than 3 of these, tracking your AI visibility is a waste of time. You are tracking a broken signal.


The three levels of control over what AI learns



You have 3 levers to pull to shape your web presence strategy. You need to manage all 3 to show up when buyers ask AI.


Your website is your baseline. But most corporate websites sound exactly the same. Everyone Has the Same Easy Button. If your site is full of generic claims about being a leading provider of solutions, AI has nothing unique to learn.


You hired smart people. Then you hid them. AI has the answers. Only your people have the stories. You need to turn that internal knowledge into public content. This is the core of Thought Leadership As A Service.


Make your content specific and useful. We worked with a client who had a single, massive page about grief. We split it into separate pages for different kinds of loss. That specificity gives AI clear, distinct concepts to learn. Give your most important feature bullets their own page.


Format matters too. We took a dense 4-page article and turned it into the Vibe Check slider. It made the information accessible. A competitor can prompt a blog post. They can't prompt your people.


If you want a metric here, don't obsess over whether one prompt mentioned you once. Watch whether your share of voice grows across the topics that matter, and whether your own pages become cited sources more often over time. That tells you the model has more to work with.


Do this first: Interview your smartest internal expert for 30 minutes and publish their exact words.


Level 2: What the rest of the web says


AI models do not trust you. They verify your claims against the rest of the internet. If your site says you are the best, but Reddit and Wikipedia have never heard of you, the AI will ignore your claims.


This is why Buyers Asked AI How Your CEO Thinks. The models pull from podcasts, industry interviews, and partner directories. Mentions, reviews, and partner pages are part of your web presence, not separate channels.


This is also where citation rate gets practical. If AI answers about your category keep citing review sites, forums, or publisher roundups where you barely appear, that is not a tracking problem. That is a market presence problem.


Do this first: Search for your brand name on Reddit and industry forums to see what the market actually thinks of you.


Level 3: Do they agree?


This is the level most marketers miss. If your site says one thing and the web says another, AI gets confused. Confusion leads to omission.


You need alignment. We saw this with a client who had a complex legal-review process for their emails. We turned those legal-review emails into a pre-check framework published on their site. Suddenly, their internal compliance language matched their public marketing language. The signal became clear.


If your homepage says one thing, partner listings say another, and reviewers use a third set of terms, your share of voice can stay flat even while you publish more content. More noise is still noise.


Do this first: Audit your top 3 partner directory listings and make sure they use the exact same terminology as your homepage.


Build, buy, or borrow your way into the conversation



Once you know what AI can learn about you, you have to fix the gaps. We call this Web Presence Intelligence. It maps where your brand shows up across traditional search, AI answers, publishers, forums, and social, compared against competitors.


Take a Fortune 25 insurer we analyzed. Their owned website content was exceptional. But when buyers asked AI about specific policy comparisons, the AI cited a massive industry forum. The insurer had zero presence there. Their competitors were actively answering questions in that forum. The insurer had a massive gap in what the web said about them.

When you find these gaps, you have to decide how to allocate your resources. You can build, buy, or borrow your way in. Stephan uses a simple framework to prioritize these marketing decisions. Most Decisions Aren't a Tattoo.


  • Hats: Easy decisions you can change tomorrow. Testing a new message on a social post. Answering a question on a forum. Try it. If it fails, take the hat off.

  • Haircuts: Decisions that take a few weeks to grow out if you mess them up. Sponsoring an industry newsletter. Buying a placement on a review site. You have to commit, but the risk is temporary.

  • Tattoos: Permanent decisions that require serious thought. Building a massive proprietary research report. Rebranding your core service lines. Measure twice, ink once.


Stop treating every AI visibility gap like a tattoo. Put on a few hats. Get a haircut. See how the models react over a few months.


Frequently Asked Questions


Can you rank in ChatGPT or other AI tools? Not in the sense of a fixed position. In SparkToro's 2025 study, ChatGPT and Google's AI rarely gave the same list of brands twice for the same prompt, and the order was even less stable. How often you're mentioned, your share of voice, and whether you appear in cited sources are more useful to watch than any single position.


What is share of voice in AI search? It is the percentage of relevant AI answers in a defined prompt set that mention your brand compared with competitors. If 100 category-level prompts are tested and your brand appears in 18 answers while 3 competitors appear in 25, 20, and 12, that gives you a directional view of who shows up most often. It is not a promise of future answers. It is a way to spot patterns.


What is citation rate in AI answers? Citation rate is how often your brand's pages, or third-party pages that mention your brand, appear in the cited sources attached to AI answers. Think of it as a clue about what evidence the model is pulling forward. A low citation rate often means the web has not produced enough trusted material about you in the places buyers already use.


Where do AI tools get information about companies? They draw on what is published about you: your own site, your profiles, and third-party sources such as review sites, forums, publishers, and reference sites. Pew's 2025 research found Reddit, Wikipedia, and YouTube were among the most cited sources in Google's AI Overviews. That's why what the rest of the web says matters as much as your own pages.


Are AI visibility tracking tools accurate? They can show patterns, such as whether you're mentioned at all for a topic, how your share of voice changes over time, or whether your citation rate improves after you publish or clean up listings. Since answers change from run to run, treat any single result as directional rather than exact. Pair tracking with a look at the sources those answers draw on.


Does Reddit or a review site really affect how AI describes my brand? It can, because AI answers pull from third-party sources you don't own. If those sources describe you differently from your website, or leave you out, the answer reflects that. Mentions, reviews, and partner pages are part of your web presence, not separate channels.


Is SEO still worth investing in with AI search? Yes. The fundamentals still apply: content that uses your buyers' language, covers their problems in depth, and is technically easy to read. Those are the same things AI tools take in. Gartner has said search engines will value content quality more to offset the flood of AI-generated content.


What is web presence intelligence? It's our approach to mapping where your brand shows up across traditional search, AI answers, publishers, forums, and social, compared against competitors. It looks at all of these together rather than one channel at a time, so you can decide where to build, buy, or borrow your way into the conversations buyers are already having.


How do I find out what AI and buyers can learn about my company? Start with the self-audit above, then get a full map of where you appear and where competitors appear instead. Our Web Presence Intelligence Report covers this for a specific topic or use case. Book a discovery call to scope one for your brand.


How we map what AI can learn about you


Wisdom in. Presence out. You cannot trick an AI model into recommending a hollow brand. You have to feed it the truth about why people trust and hire you.


We help service firms and expert-led brands figure out exactly what the internet knows about them. Through Web Presence Intelligence, we map your footprint across traditional search, AI answers, forums, and publishers. We find the gaps. Then we help you extract the stories from your smartest people to fill those gaps.


Help, not a pitch. If you are tired of tracking numbers that change every Tuesday and want to build a footprint that actually drives trust, Book a Discovery Call. Let's talk.

Comments


bottom of page