Your AI Visibility Isn't One Number: Why ChatGPT, Claude, and Gemini Cite Completely Different Sites

Here is a frustrating thing that happens to a lot of businesses right now. You ask ChatGPT to recommend a tool in your category, and there you are. You ask Gemini the same question, and you have vanished. You try Claude, and a competitor you have never worried about is the one getting named.

Same business. Same question. Three different answers.

Most people assume this is randomness, or that the models are just fuzzy and unpredictable. It is neither. There is a real structure underneath it, and once you understand that structure, the whole idea of a single "AI visibility score" starts to look misleading.

Different assistants read from different libraries

The simplest way to think about it: each AI assistant is reading from a different library when it goes to look something up.

When an assistant answers a question that needs fresh, factual, web based information, it does not invent the answer from memory. It retrieves pages from a search index and grounds its response in what it finds. That retrieval step is the part people overlook, and it is the part that decides whether you get cited.

The catch is that the assistants do not share one index. They each lean on a different underlying search engine:

• ChatGPT's web answers lean heavily on the Bing index.
• Google's AI Overviews and AI Mode ground on Google's own index.
• Recent data suggests Claude's visibility tracks closely with Brave Search rankings.

So when you "win" in one assistant and lose in another, you are not seeing model randomness. You are seeing the difference between three search libraries that rank pages differently. If you rank well in Bing but poorly in Brave, ChatGPT may love you while Claude has never heard of you. That is not a bug in the model. It is a direct reflection of which index it reads.

The data backs this up, and it is more extreme than you would guess

This is not a hunch. One study ran 1,860 queries across major AI models and found that 72.8 percent of businesses were known by only a single AI ecosystem. Read that again. For nearly three quarters of businesses, being recommended by one assistant told you almost nothing about whether any other assistant would recommend them. Only about one in nine businesses showed up consistently across all the major models.

If you have been treating "AI visibility" as one thing you either have or do not have, that number should stop you. Most businesses do not have AI visibility in general. They have visibility in one corner of the AI world and a blind spot everywhere else, and they usually do not know which is which.

Why a single AI visibility score hides the thing you need to know

Plenty of tools will hand you one blended "AI visibility" number. It looks tidy. It is also the wrong unit of measurement, for the same reason a single "search visibility" number that averaged Google and Bing together would have been useless ten years ago.

A blended score buries the only insight that is actually actionable: where you are winning and where you are invisible. If your number is "62 out of 100," you have no idea whether that means you are strong in ChatGPT and absent from everything else, or evenly mediocre across the board. Those two situations call for completely different work. The first means you have a Brave and Google problem to solve. The second means you have a foundational content and authority problem. One number cannot tell them apart.

The volatility makes it worse. Tracking studies have found that anywhere from 40 to 60 percent of the sources cited in AI answers change from month to month. A single score that bounces around every month looks like noise and gets ignored, when the real signal, which engine cites you and which one drops you, is sitting right underneath and never gets surfaced.

What to measure instead

If AI visibility is really several different visibilities stacked on top of each other, then the way to measure it is per engine, mapped to the index each one reads.

That means checking your presence separately across ChatGPT, Claude, Gemini, and the others, and reading each result in the context of the search index behind it. Showing up in ChatGPT but not Claude is a Bing strength and a Brave weakness. Showing up in Gemini but not ChatGPT points you at the gap between Google's index and Bing's. The per engine view turns a vague "you have low AI visibility" into a specific, fixable list.

There is a quieter implication here too. Because so much of AI citation flows downstream from these search indexes, a lot of your AI visibility is still earned through fundamentals you already understand. Ranking well in the underlying search engines, having content an assistant can actually retrieve and extract, and being genuinely worth citing all carry over. AI search did not throw out everything you knew. It just spread it across more libraries.

The takeaway

There is no single "rank number one in AI" to chase, and there is no single visibility score that captures the whole picture. Each assistant reads from its own library, and most businesses are known in one library and invisible in the rest.

So the next time you check whether AI recommends you, do not ask "am I visible in AI." Ask "which assistants cite me, which ones don't, and what does that tell me about where my work needs to go." That is the question with an answer you can act on.

If you want to see this split for your own site, Potatometer checks your AI search readiness across engines so you can see where you show up and where you have a blind spot, instead of one number that hides both.

← Back to Blog