I know a lot of people are trying to create tools that track brand mentions across the different AI search tools. I applaud their efforts and encourage them to keep working on it, but not for the reasons you think. It’s not because I think they can create a tool for that right now. it’s because they will be ready when there is a way to do it reliably.
Industry data from 2026 confirms what I have been saying from the beginning
The “Segment of One”: Modern AI uses “dynamic profiles” that update in real-time based on the user’s browsing behavior, scroll depth, and even device type. This makes a single “rank” or “mention score” for a brand technically impossible to standardize. This is not like SEO or ranking in Google where there is actually a database to draw from. The difference between measuring AI visibility and Google rankings could not be further apart.
Synthetic Data Problem: Tools like Semrush and Nightwatch mostly rely on synthetic queries.These are automated bots asking a fixed set of questions via API. These bots do not have the user’s context, history, hobbies, or professional, which is why their “visibility scores” mostly differ from what real users see.
The “Transparency” Gap: OpenAI, Google, Anthropic & Perplexity AI do not provide analytics APIs for brand mentions. Until they do, any tool claiming to “track” mentions is effectively using “directional surveillance” or simulations rather than hard data.
As much as we all wish we could track brand mentions and produce actionable data, it isn’t happening yet. In my opinion, it will. That’s why I still encourage people to keep learning, researching and keep building tools.
But, and there is always a but. If you see anyone, including Ahrefs, SemRush or anyone on TikTok, YouTube and Instagram claiming they can track this for you, then it is misleading at best, scamming at the worst. Someone saying it might actually believe what they are saying. It doesn’t matter because the bottom line is that it cannot be done yet.
How Google AI Search Works vs How Other AI Search Tools Work
Google AI Search retrieves information by pulling “grounded” data from its existing search index and passing it through an LLM to summarize. It specifically uses Retrieval-Augmented Generation (RAG) to ground the LLM in that index, which is why citing high-authority sources is its primary surfacing mechanism. It identifies what it considers the most relevant snippets from those pages and displays them as citations within an AI Overview. Unlike standard search, it isn’t just looking for a page to rank; it’s looking for the most reliable facts to build a direct answer for the user.
GEO / Ai Search Optimization focuses on how AI models extract and reformat information during the generation process. It retrieves data by scanning for high-density answer cues, such as structured lists, tables, and authoritative quotes that are easy for a model to digest. It surfaces results by weaving these specific content blocks into the AI’s generated response, prioritizing answer-first formatting that fits perfectly into a chat-style interface.
The Silver Lining: Why These Tools Still Matter
While you can’t track every mention, these tools can show you where an AI is consistently citing a competitor or a low-quality source like an old Reddit thread or a hallucinated answer instead of your current, authoritative content. This tells you exactly where your Entity Footprint is weak.
Learn how Chris McElroy SEO Agency is helping clients strengthen your brand’s entity
If multiple tools start showing that different AI models are describing your brand as expensive or outdated, it doesn’t matter if the prompt was perfectly personalized, it means the underlying training data or RAG (Retrieval-Augmented Generation) sources have a bias against you that needs fixing. The bias isn’t personal the way that sounded. And it is possibly simple to fix. AI might not be able to find the information and messaging you want it to find.
Entity Optimization is the key to being mentioned in AI search tools AND in Google. When you focus just on the surface SEO advice out there or on GEO advice, then you’re only scratching the surface. You just need to help AI-powered search and Google’s Knowledge Graph define it.
Everything relevant and consistent connected to your brand, your products, who you are, your profiles, and more becomes part of creating your brand as an entity. Entity Optimization is about helping AI understand and trust your brand.
Enlarge your thinking beyond “I targeted x keywords and wrote content and got backlinks so it should rank well in Google and AI” or “I put in an FAQ and schema markup so Google should use something there as a rich snippet and AI should surface those answers to users”
There is not a formula to follow. Content Targeting X + X Number of Backlinks from X DA domain names = success. This formula has gotten results in Google in the past and that is not a sign that Google is doing a good job. It’s an exploit they failed to solve.
Entity Optimization And Why It Matters
Entity Optimization or Entity SEO, is the process of establishing your brand as a verifiable node within a Knowledge Graph. It retrieves information by looking for established relationships between your brand and specific topics, rather than just matching keywords. It surfaces your brand because the AI understands who you are and trusts you as a primary source of truth, making you the “default” mention for a specific category.
And who you are is just the start. AI wants to understand the 5 Ws, the who, what, where, when and why of you personally, your relationship to the business name, the business name to the brand names you use, the products that fall under those brand names and even the slogans you use. Those are each entities on their own. AI wants to know the relationship between them. And if you add How to the 5 Ws you have something solid.
In addition to that, AI wants to know exactly who your products or service are for, your ideal customers. This part is counterintuitive, but it also wants to know who your services and products are NOT designed for. AI will trust you much more if both of those are defined.
This is the infrastructure layer that provides a machine-readable roadmap for AI crawlers. Ideally it will retrieve data via direct instructions found in discovery files like llms.txt and ai.json, which should tell the bot exactly which parts of your site are authoritative.
Important to note: There is no conclusive data that either of these files influence AI search results. llms.txt has been proposed as a standard but there is no evidence that proves any of the AI search tools pay attention to it. Showing that it got crawled does not equal that it made any difference at all.
Again, ideally it will surface results by removing the guesswork for the AI, ensuring it uses your verified data instead of hallucinating or relying on outdated third-party mentions. My take is this: It doesn’t hurt to have those files on your server. Just don’t expect it to be some magical cure. Even Robots.txt is just a suggestion, not a rule, and it’s been established for years.
Is anyone here actually seeing a crawl-to-citation correlation with their llms.txt files yet?”
I welcome your questions and your comments, even if you disagree. If you have something to add to make this a valuable discussion, please add them. We are all still learning about this.
Technical Infrastructure And The Rumors About Server Files Like LLMs.txt
Optimizing for AI is part of doing SEO. However, thinking that it is the same as optimizing for Google is wrong. Thinking that AI just gets it’s results from search engines is wrong. We have to quit trying to get our heads wrapped around AI search optimization by thinking in SEO analytics terms.
- “Well, robots.txt works so that means llms.txt work too.” It doesn’t mean that at all. This is just an SEO mindset that may or may not ever happen.
- “You have to have schema markup that passes the Google Rich Snippets Test if you want to get mentioned in AI.” No you don’t. That doesn’t mean not to use schema markup because it does help AI understand the content on your page. However, AI can see everything in the source code of your web page and can pull answers from it, if it’s well structured and clear. Add to that, other AI search tools do not care what Google will consider for rich snippets. There are over 800 types of schema you can use.
- “Just put FAQs in your page and those answers will be used by AI.” Again, no. For one, Google isn’t going to pay attention to them unless your website is a university, government, or another highly trusted source. Mostly because of FAQ spam. That’s the other thing. If you use an FAQ it should be for your readers, not for SEO or AI SEO. If your answers are salesy, contain a bunch of links or brand mentions, it’s worthless and will get ignored.
Your infrastructure layer provides a machine-readable roadmap for AI crawlers. Ideally it will retrieve data via direct instructions found in discovery files like llms.txt and ai.json at some point in the future, and that would tell the bot exactly which parts of your site are authoritative.
Important to note: There is no conclusive data that either of these files influence AI search results. llms.txt has been proposed as a standard but there is no evidence that proves any of the AI search tools pay attention to it. Showing that it got crawled does not equal that it made any difference at all.
Hopefully in the future, it will surface results by removing the guesswork for the AI, ensuring it uses your verified data instead of hallucinating or relying on outdated third-party mentions. My take is this: It doesn’t hurt to have those files on your server. it doesn’t hurt to have legitimate FAQs that are helpful to your visitors. You should have been using schema markup before AI search was a thing and using it now. Just don’t expect those things to be some magical cure. Even Robots.txt is just a suggestion, not a rule, and it’s been established for years.
Are you actually seeing a crawl-to-citation correlation with their llms.txt files yet?” I welcome your questions and your comments, even if you disagree. If you have something to add to make this a valuable discussion, please add them. We are all still learning about this.
TLDR Summary
- The Tracking Myth: Current tools for tracking AI mentions are directional at best. Because AI responses are personalized (“Segment of One”) and official APIs don’t exist, a universal “rank” is technically impossible right now.
- Entity > Keywords: Stop chasing surface-level SEO. AI surfaces brands it understands and trusts within a Knowledge Graph. Your goal is to become a verifiable “node” by defining your identity (the 5 Ws) and who your product is not for.
- The Strategy: Success in AI search is a mix of RAG-grounding (Google), GEO (formatting for extraction), and Technical Infrastructure (using files like
llms.txtandai.jsonfor deterministic discovery). - The Bottom Line: Don’t wait for perfect data. Start building your Entity Footprint now so that when transparency arrives, the AI already trusts you as a primary source of truth.
