What AI visibility actually means

To appear in AI answers, a company has to be findable, described the same way everywhere it is described, corroborated somewhere other than its own website, and relevant to the question the buyer actually asked rather than to the name of its category.

There is no submission form, no directory to join and no switch to turn on. Answers are assembled from material these systems can retrieve and weigh, so the work is making sure there is material, that it agrees with itself, and that it addresses the question.

And appearing is not a single yes or no. A company can be named often and described wrongly, named last in every answer, or named with nothing supporting the claim. Those are four different results with four different fixes, which is why one number is not enough to work from.

Is there a way to submit my company?

No, and it is worth saying plainly because it is the first thing most people look for.

None of the six systems takes submissions, sells inclusion or keeps a supplier list you can apply to join. Anyone offering to add you to an AI's index is selling something that does not exist. What does exist is the ordinary, slow work of being easy to find, easy to understand and confirmed by somebody other than yourself.

You cannot buy your way in. You can only be worth including.

Appearing is not one thing

Four separate readings, and a company can pass one while failing another.

AI Visibility is how often you appear at all across the questions your buyers ask. AI Recommendation is how often you are put forward as an option worth considering rather than merely mentioned. AI Position is where you sit among the suppliers named alongside you. Entity Accuracy is whether the description of you is correct. Underneath the first of them sits citation share — the share of tested questions in which your company is named, which is a component of AI Visibility and not another word for it.

All five are defined on Baseline, the step of our method that produces them. The reason to keep them apart is practical: being absent, being last, and being described as something you stopped making in 2019 are three different problems, and the work that fixes one does nothing for the others.

What makes a company appear

Four things, in the order they have to happen.

It has to be retrievable. A page that is not crawled, not indexed, or too broken to process cannot be drawn from. This is unglamorous and it is the precondition for everything else.

It has to answer the question. Not the category term — the question. "Who supplies case packers for glass jars at low volume" is answered by a page about that, not by a page headed "Case packers". One buyer question is also searched as several narrower ones, which is set out in query fan-out.

It has to agree with itself. One description of your company, the same on your site, in directories, in listings and in your structured data. Contradiction is worse than silence, because there is no settled version to use.

It has to be confirmed elsewhere. Your own site says what you say about yourself. Independent sources are what make that checkable, and they are usually the thing a named competitor has and an absent one does not.

Appearing and being recommended are different

Being mentioned in passing is not the same as being put forward.

An answer can list your company among ten in a sentence about the market, and name three suppliers when the buyer asks who to approach. You were visible in the first and absent from the shortlist in the second. The measurement that matters depends on which question was asked, which is why a question set is agreed before anything is tested rather than after.

You can appear and still lose

Three ways, all common.

Named last, every time. A buyer reading three names contacts the first two more often than the third, and a consistent third place is a finding rather than a success.

Named for the wrong thing. Known as a general supplier when the buyer asked about a specific application, or described by a product line you discontinued years ago. That is an accuracy problem and it is usually the fastest of all of these to fix.

Named with nothing behind it. An answer that mentions you but rests on no source is a weaker mention than one that cites a trade body, and it tends not to survive the next time the question is asked.

If enquiries have quietly fallen while your rankings held, that is the same story told from the other end — your B2B website stopped producing enquiries.

How would you check?

Write down thirty questions your buyers ask before they know who to call. Put each to ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity and Claude, in a logged-out session. Record which companies are named, in what order, and what sources the answer cited.

Then read your own entries twice: once to count how often you appear, and once to check whether the description is correct. Most companies find the second read more uncomfortable than the first.

What this article cannot tell you

It cannot tell you what a good number looks like. There is no published benchmark for any of these measures in industrial equipment, because nobody has produced one — which is why we are building the AI Visibility Index rather than quoting a figure we would have had to invent.

It cannot tell you how the systems weigh anything. They do not publish it, and any firm claiming to know the weighting is guessing confidently.

And a single reading is a snapshot. AI systems are probabilistic and change continuously, so what tells you anything is the same questions, asked the same way, over time.

What to do next

Test before you plan. A definition of the discipline is a poor substitute for knowing whether your own name comes back.

Thirty questions and an afternoon will tell you that. If you would rather have it measured properly — agreed questions, all six systems, your named competitors alongside you, every source recorded — that is the AI Visibility Audit.