Can AI find you?

When buyers ask AI about your category, someone gets named. Bellwether puts 22 real buying questions to ChatGPT and Gemini, then shows you whether it is you, and who it is instead.

What is AI visibility?

AI visibility is how often, and how accurately, an answer engine names your business when someone asks a buying question in your category.

It is not the same thing as search ranking. A page can sit at the top of Google and never once be named by ChatGPT, because an answer engine does not return a list for the user to choose from. It synthesises a single recommendation, cites two or three sources, and everything else is invisible. There is no page two to be on.

Which means the question is no longer whether you can be found. It is whether you get named.

How answer engines decide who to name

Four things determine whether you show up in an answer. Every gap the check finds falls into one of them.

A page answers the question as asked

Not the topic, the question. “Golden Circle tours” is a topic. “A Golden Circle tour I can book online with instant confirmation” is the question, and it needs a page that answers it in those words.

The specifics are stated, not implied

Free cancellation, group size, price range, service area, turnaround. An engine will not infer what you did not write. Competitors who state it get named for it.

Someone else corroborates you

Directories, reviews, press, partner pages. Engines weight claims they can find in more than one place, which is why aggregators are so hard to displace.

Machines can actually read it

Schema markup, a working llms.txt, and robots rules that do not quietly block AI crawlers. The most common single failure, and the fastest to fix.

This page is built the way we would tell you to build yours: a direct definition near the top, stated specifics, a marked-up FAQ, and a readable llms.txt. If the method did not work we would be hard to find too.

22 questions. Two engines. Forty seconds.

We do not ask the engines about you. We ask them the questions your buyers ask, then look at who gets named, which is the only measurement that matches how a purchase actually starts.

The check runs against ChatGPT and Gemini. Two engines is deliberate: they are the most widely used, and a question you lose on both is a real gap rather than one model's quirk. It is a baseline, not a census.

Discovery questions

“Best [category] in [place]”, where an engine builds its shortlist.

Qualifier questions

“…with free cancellation”, “…under $500”, “…for small groups.”

Comparison questions

“X vs Y”, “specialist or aggregator”, where you are defined by others.

Direct-brand questions

“Is [you] any good?”, what an engine says when asked outright.

Where can I find a Golden Circle tour I can book online with instant confirmation?

Named: GetYourGuide, Guide to Iceland

Which Iceland tour companies offer free cancellation if the forecast is poor?

Named: three local competitors

What is the advantage of a specialist over a large aggregator?

Named: nobody in this category

Small-group Iceland day trips leaving from Reykjavik?

Named: you, 4th of 6

Question set shown is from a real scan of an Iceland tour operator, used here as an example. Yours is generated from your own site and category.

A score, the gaps behind it, and what to publish

The score and the first two gaps appear on screen straight away, with no email and no account. The full ordered plan is the only thing we ask for an address to send.

Named 4 times. Your competitors, 32.

Across 22 buying questions this operator was named four times. Three aggregators absorbed the rest. Every gap below is a page that does not exist, or one that does not answer.

Who gets named instead

No Golden Circle page with instant confirmation

Two aggregators are cited here because both state online booking and instant confirmation on the page itself. Yours states neither.

Where can I find a Golden Circle tour I can book online and get instant confirmation for my dates?

Nothing answers “specialist vs. aggregator”

Buyers ask this directly and no one in the category answers it, so the aggregators define the comparison unopposed. An open goal.

What is the advantage of booking through a specialist rather than a large aggregator?

+ 7 more gaps in the full plan, ranked by impact, each with the question it wins and the page that should answer it.

How to fix each kind of gap

Four gap types, in the order they usually pay off. You can do all of this yourself. Bellwether exists to tell you which ones matter and to write them faster, not to gatekeep the method.

Machine-readable files

An afternoon. Often the single biggest jump.

Check three things in order. Your robots.txt, because a surprising number of sites block GPTBot or Google-Extended, the crawlers behind the two engines this check tests, without anyone having decided to. Worth clearing ClaudeBot and PerplexityBot at the same time. Your schema markup: Organization, Product or Service, and FAQPage at minimum, so claims arrive as structured facts rather than prose to be parsed. And llms.txt at your root, a plain-language summary of what you sell, to whom, and where, which is the cheapest way to hand an engine an accurate description of you. None of this changes what a human sees. All of it changes what a machine can state about you with confidence.

One page per real question

A week per page, and they compound.

Take the exact question from the report and make it the page. Put the answer in the first paragraph, not after a scene-setting introduction, then support it with the specifics an engine can quote: the booking method, the confirmation time, the cancellation window, the price band, the group size. A page that answers in its first fifty words gets cited. A page that warms up for three paragraphs gets skipped. Write for the person who asked the question, and the engine follows. The reverse does not hold.

Define the comparison yourself

Usually the emptiest ground.

Buyers ask engines to compare before they ask anyone else, and most categories have nobody answering, so the largest aggregator's marketing becomes the default answer. Write the honest version: where you genuinely win, where the other option is the better choice, and who each suits. Engines cite balanced comparisons more readily than one-sided ones, and so do people.

Write down what you assume is obvious

The cheapest fix on the list.

Most businesses lose qualifier questions on things they already do. You offer free cancellation, you cap groups at twelve, you cover the whole metro area, you turn work around in two days, and none of it is written anywhere a machine can find. Add the sentence. It is often the difference between being named and being absent on a question you would win outright in person.

The check is one of over fifty tools

AI Presence is where most people meet Bellwether. It sits inside a system of more than fifty tools that watches your market daily and turns what changes into work you can actually do.

Opening Bell

A daily brief on your category: competitor moves, sentiment shifts, search and market signals, with the two or three that matter ranked at the top.

Competitor Intelligence

Messaging, positioning and content gaps for the companies being named instead of you, kept current rather than audited once a year.

Social & Review Sentiment

What people say about you where engines can read it, the corroboration layer that decides whether your claims get believed.

Topic Generator & Creator

Every gap becomes a brief, and every brief becomes a draft, in your voice, grounded in the scan that found it.

Content Plan

Gaps land on a calendar in impact order with the source question attached, so the work has a sequence instead of a backlog.

Re-scan & track

Run the same 22 questions monthly and watch the score move, the only way to know whether any of the work landed.

The check is free. Keeping score is not hard either.

You can take the plan and never come back, it is yours. If you would rather watch the number move, a free account re-runs the scan monthly.

The check

Everything on this page, for one site.

Fix it faster

The intelligence layer plus the writing tools that close the gaps.

Many brands, one view

For agencies and multi-brand teams tracking visibility across a portfolio.

Before you run it

If something is not here, the check itself will probably answer it faster than we can.

Yes. The score, the competitor ranking and the first two gaps appear on screen with no email and no account. We ask for an address only to send the full ordered plan, and you can take that plan and never come back.

An SEO audit tells you how a page is likely to rank in a list of links. This tells you whether an answer engine names you when a buyer asks a question out loud. The two overlap, because good structure helps both, but a site can rank well and still be absent from every AI answer in its category, which is the gap this measures.

ChatGPT and Gemini. They are the two most widely used answer engines, and together they give a dependable baseline. If neither names you on a question, that is a real gap rather than one model behaving oddly. Engines answer differently and change their answers often, so the check reports each one separately instead of blending them into a single number. It is a baseline, not a census.

They are generated from your site and category, then spread across four intents: discovery, qualifiers, comparisons and direct brand questions. You can see every question in the report and swap any of them before a re-scan.

Machine-readable fixes such as robots rules, schema and llms.txt can register within weeks. New answer and comparison pages take longer because they need to be crawled and, ideally, corroborated elsewhere. We would rather you re-scan monthly and see something real than promise a timeline we cannot control.

A plain-text file at the root of your site giving AI systems a clear summary of what you do, what you sell and who you serve. It is an emerging convention rather than a settled standard, so it is not a magic switch, but it costs an afternoon and removes any ambiguity about how you describe yourself.

We keep the scan so you can compare it against later ones. If you never create an account, the report expires on its own. Nothing is sold or shared, and you can ask us to delete it at any point.

Can AI find you?

You will know before you finish your coffee, and you will know exactly who is being named instead.