AI visibility, and how to tell if you show up in AI answers

A buyer asks ChatGPT who they should hire, or types a question into Google and reads the AI summary instead of scrolling. You want to know one thing: does my business show up? The advice online does not help much. Most of it is written for national brands tracking hundreds of prompts, or it is vague filler that says "get mentioned by AI" without explaining what that means or how to check. This is the practical version, for a business that needs to bring in work.

AI visibility is not one thing

The first problem is the word itself. Asking "are we visible in AI" is really four separate questions, and they do not move together: what the model knows about you by default, whether it cites your page as a source, whether it says your name to the user, and how accurately it describes you.

These come apart in the data. Analyst Kevin Indig's study of how often AI names the brands it cites found that AI tools cite a source far more often than they say the brand behind it: across thousands of appearances, only about one in seven turned into both a citation and a named mention. Your page can be doing the work while your name never reaches the reader. So when someone says "track your AI visibility," ask which of the four they mean. A dashboard that reports one score is hiding the question that matters.

How to check whether you show up

You can do this yourself for free, and you should, before paying anyone. The method is simple: build a set of real customer questions, run them through the engines the way a customer would, and log not just whether you appear but whether you are linked and how you are described.

1. 01

   List the questions a real customer would ask

   Ten to fifteen, in their words, including local ones like "best [your service] in [your city]" and "who should I hire to do [job] near me."

2. 02

   Run them through the engines

   ChatGPT, Perplexity, Gemini, and a normal Google search where you watch for the AI summary at the top.

3. 03

   Log the three things that matter

   For each answer: do you appear at all, are you linked, and how are you described.

4. 04

   Run it logged out and located

   Use a private window with location set as close as possible to the area your customers are in, not your own desk. Logged in at your own address, the AI may show you a personalized version nobody else gets.

5. 05

   Run the set more than once

   The results move between runs, so a single check is misleading. Track how often you appear, not a one-time position.

That last point is the one most people get wrong. SparkToro's research with Gumshoe.ai ran nearly three thousand recommendation prompts across ChatGPT, Claude, and Google's AI answers and found the same question returns the same list less than one percent of the time, and the same list in the same order only about once in a thousand runs. The brands, the order, and the length of the list shift on almost every run. The honest takeaway is that how often you appear across many runs is a real measurement, and your rank in ChatGPT is not. If a tool or agency sells you a fixed AI ranking position, they are selling you noise. Track appearance frequency over time instead.

Why your result keeps changing

The reason the list moves is structural, not a glitch you can fix. AI engines rotate a large share of the sources they cite every week. But underneath that churn sits a small, stable core, the few names a model returns again and again, and that core is what the work aims at.

SISTRIX studied more than eighty thousand prompts over seventeen weeks and found AI engines rotate a large share of their cited sources every week, with ChatGPT replacing around three quarters of them, and no sign of settling down. The same study found the useful part: across most prompts there is a small, stable core of sources that appear week after week while everything around them rotates. So the goal is not to win a citation slot that vanishes by Friday. It is to become one of the names the model reliably associates with your kind of work in your area. That is a reputation and consistency game, and it is steadier and slower than the dashboards suggest.

What moves AI visibility

The work splits into two layers, and most owners focus on the first. On your own site, the foundation decides whether you are eligible at all. Off your site, where you are mentioned decides whether the engines trust you enough to repeat your name. The second layer is the one most owners underrate.

On your own site, the foundation still decides whether you are eligible at all. On Google specifically, AI features are tied to the same index and ranking systems as normal search, so if a page cannot be crawled, understood, or ranked, it is unlikely to be pulled into the AI answer. Past that, the engines favor content they can use cleanly: a direct answer near the top, self-contained passages, and real specifics rather than slogans. Concrete detail and genuine expertise travel; thin filler does not.

The second layer surprises people, and it matters more than most expect: where you show up off your own site. Ahrefs studied seventy-five thousand brands and found that unlinked mentions across the web tracked with AI visibility more than three times as strongly as backlinks did. The models read who you are from how often your name comes up in normal writing, in reviews, in threads, in roundups, not from a link graph. Being talked about on the sites the engines trust feeds the answer even when nobody links to you. For many small businesses this off-site presence is the missing lever, and it is the one most on-page checklists barely cover.

If you run a local business, most AI visibility advice misses the point

Almost everything written about AI visibility is built for national brands competing on broad topics. A local, owner-operated business plays a different game, and the recent evidence is specific about it: the bar is high, but the signals that clear it are exactly the ones a focused operator can control.

The bar is high. SOCi's 2026 study of more than three hundred thousand locations found ChatGPT recommended barely over one percent of them, compared with about a third in Google's traditional local pack. But the signals that earn that recommendation are controllable. The annual Whitespark local search ranking factors survey found that, for AI search specifically, the weight on Google Business Profile drops while citations and mentions across the web rise. Sterling Sky, testing thousands of businesses, found that recent reviews matter more than lifetime count, and that reviews with written text matter more than star-only ratings because the engines read the words to understand what you do. The same work found that hiding your address as a service-area business correlated with worse visibility, even though Google's own guidance often points service-area businesses toward hiding it.

There is also a blind spot worth naming. ChatGPT's web search relies on Bing, not Google, and Bing-fed local results lean on sources many owners ignore, including Facebook reviews and a verified Bing Places listing. A business that has optimized only for Google can look fine in Google's AI summary and be invisible inside the most-used assistant. Verifying Bing Places and tending Facebook reviews is a separate task most owners never hear about.

This is familiar ground for the studio. When Jardine Studio built the local foundation for Palisades Lodge, a destination property, part of the work was a complete, accurate profile and review growth from a low baseline. Those are the kinds of signals the research now shows feeding local AI recommendations. No one can promise a given AI engine will name a business. But the entity work that has always supported local search now supports local AI answers too.

The honest limits

A few things are worth saying plainly, because the people selling AI visibility usually will not. Nobody can guarantee a citation, the measurement is genuinely imperfect, and a fair amount of the popular checklist is theater. Knowing which is which saves money.

Nobody can guarantee a citation. AI answers are generated probabilistically and the sources rotate weekly, so guaranteed placement is not something an honest provider offers. The work improves the odds; it does not control the output.

Measurement is genuinely imperfect. AI tools often strip the referrer, so the customers they send you can land in analytics as "direct" traffic. Many owners may already be getting AI-referred visits they cannot see, then conclude AI does not matter. The first step is a simple analytics setup that groups AI sources before you pay for a tracking tool.

A fair amount of the popular checklist is theater. Google has said on the record that the llms.txt file some guides push is not used by its systems, with John Mueller comparing it to the keywords meta tag search engines abandoned long ago. Stuffing pages with FAQ markup is not a strategy either: when Ahrefs tracked nearly two thousand pages that added schema, AI citation rates barely moved. If FAQ content is useful and visible, mark it up; if it is only there to chase citations, it is noise. Be wary of anyone selling a brand-new acronym with a lot of urgency. The louder the new jargon, the more likely it is noise.

Where that leaves you

AI visibility is four questions, not one: what the model knows, whether it cites you, whether it names you, and how it describes you. Check it yourself, track how often you appear instead of chasing a made-up rank, fix the foundation on your own site, and put real effort into off-site and local signals: reviews, profiles, citations, mentions, and being named in the places the engines trust.

If you want a read on where your site stands across search and AI visibility, the Free SEO Audit crawls the site and flags the gaps. The deeper work of structuring a site to be easier for AI systems to understand, summarize, and cite is what the studio runs as AI search optimization. For how the two disciplines fit together, see GEO vs SEO, and what it changes for your site; for the on-page craft, what passage optimization actually means.

FAQ