If you investigate AI search visibility, you need to know which businesses assistants recommend for your buyers' actual requirements. Ask an assistant the questions your buyers ask and record which companies come back. The list is frequently not the one you would have written, and the difference is the useful part.
Why the named set differs from your competitor list
Companies build their competitor list from deals they lose and from firms that resemble them. An assistant assembles its list from what has been written about the market.
Those diverge for understandable reasons. A competitor with a strong content operation appears in answers regardless of market share. A large incumbent appears because it is written about. A company you compete with constantly may be absent because nobody covers it.
So the named set describes who is present in the material your buyers encounter, which is a different and equally important question from who you lose deals to.
Running it
Use category and problem questions rather than brand questions, since brand questions return the brand you named. Ask the way a buyer would: what tools do X teams use for Y, how do I solve Z, what should I look at if I need W.
Run them across providers and over several weeks. A single run is one observation of a system that varies, and a name appearing once is not evidence of an established position.
Include persona variants where your market divides. The set named for a small buyer and a large one often differs substantially. That difference tells you which end of the market each competitor owns in the written record.
Reading the result
Four categories come out, and each means something different.
| What you see | What it means | What to do |
|---|---|---|
| Expected competitors, consistently named | The written record matches your understanding | Confirm they are in your campaign's competitor list |
| A company you do not consider a competitor | Buyers are being pointed at it for questions you answer | Find out why it is being named. Usually content or a well-cited third-party page. |
| A competitor you take seriously, absent | They win deals without a presence in the written record | Useful intelligence. Their absence is an opening for you. |
| Your own absence | The material does not associate you with the question | Work from the sources those answers cite |
The second row is the finding that justifies the exercise. A company you do not track is being recommended to your buyers, and you would not have known.
Investigating an unexpected name
When a company appears that you did not expect, the question is what is producing it.
Look at the sources the answers cite when naming them. Usually one of three things is happening: they appear in a widely cited roundup or comparison, they have published material that directly answers the question, or a third-party review site covers them prominently.
All three are reproducible. That is the point of looking: an unexpected competitor's presence is normally traceable to specific pages, and those pages are places you can also appear.
Update the campaign's competitor list
Add the relevant names to your campaign's competitor list. Because labels are applied when a report is generated rather than when data was collected, adding a competitor changes how your existing history reads without new collection. You can see where they have been appearing.
Adding them also improves the link gap analysis, since a company named in answers is a company whose placements are worth comparing against yours.
Do not add every name. A list of fifteen competitors produces a weak signal everywhere, and a link gap against fifteen sites returns rows where a shared link means very little.
What this does not tell you
It does not measure market share, revenue or customer satisfaction. A company named often may be small and well written about.
It does not explain why an assistant selected those names, which is not observable from outside.
And it is not stable. Answers vary between runs, so a name appearing in one collection and not the next has not necessarily changed position.
When reviewing the competitive set, investigate unfamiliar recommendations in their original answers. You can then decide which businesses belong in ongoing research.
See LLM Top Sites, monitoring competitor mentions, or how personas change the answer.