AI visibility is the share of AI answers in which a brand appears, for the questions that brand's customers ask. "Appears" covers two things: being mentioned by name in the text of the answer, and being cited as a source through a link to one of the brand's pages. A brand with high AI visibility is named or linked in most answers to its core questions on most platforms. A brand with low AI visibility is absent, even when its own pages rank well in traditional search.
The term exists because AI search changed what "ranking" means. There is no position one in a written answer. There is a set of names the model chose to include, a set of pages it chose to cite, and an order and tone it gave them. AI visibility is the measurement of those choices.
What does AI visibility consist of?
Most tracking tools break the idea into five parts.
Mentions. Whether the answer names the brand at all. This is the basic unit, and it is independent of the brand's own website: the model can name a company it read about on a review site without ever citing the company's pages.
Citations. Whether the answer links to a page on the brand's domain. Citations send traffic and show that the brand's own content was retrieved and used. How AI search chooses which sources to cite explains what makes a page citable.
Position. Where in the answer the brand appears. The first name in a list of recommended options is seen by more people than the fifth, and some answers mention a brand only in a caveat.
Sentiment and context. What the answer says about the brand: recommended, listed with reservations, or mentioned as the thing to avoid.
Share against competitors. The same measurements for the other brands named in the same answers. Visibility is relative; a brand mentioned in half of the answers is doing well if its competitors are mentioned in a tenth, and badly if they are mentioned in all of them.
Why can't AI visibility be read from one check?
A single AI answer is one sample from a distribution. Ask the same question again and the answer changes, for three reasons.
The model samples its words, so two runs of the same prompt produce different sentences and sometimes different lists. The retrieval step can return a different set of pages, because the index updates and because the model may write different search queries. And the platforms themselves change: a model update or a change in the retrieval pipeline shifts answers across every topic at once.
Because of this, a brand that appears in one answer today may be absent tomorrow without anything changing on its side. Reliable numbers need the same prompts run repeatedly, across several platforms, over weeks. Why do AI answers change between runs? goes into the sources of variation, and the Searcherries article on measuring AI visibility shows how much a single check can mislead.
How is AI visibility measured?
The method is the same whether it is done by hand or by a tracking tool.
- Choose the questions. A prompt set of ten to fifty questions customers actually ask, phrased the way they ask them ("best project management tool for a small agency") rather than as a keyword ("project management software"). The set should cover the brand's categories, comparisons with competitors and the problems the product solves.
- Choose the platforms. ChatGPT, Perplexity, Google AI Overviews, Gemini, Claude and Grok answer differently and draw on different sources, so a brand's visibility is usually uneven across them.
- Run the prompts on a schedule. Daily or weekly, with the answers stored in full.
- Record what each answer contains. Which brands are named, in what order, with what sentiment, and which URLs are cited.
- Compare over time and against competitors. The useful outputs are a share of answers that mention the brand, the same share for each competitor, the list of pages that get cited for each question, and the change in all of these from one period to the next.
The cited pages are often the most actionable part of the data. They show which review sites, forums and comparison articles the platforms trust for a topic, which is where a brand needs to be present to be named.
How is AI visibility different from search rankings?
Rankings measure one page for one query in one engine, and the result is a number that holds until the next crawl. AI visibility measures a brand across many answers, and the result is a share that moves with every run.
The sources differ too. A ranking depends on the page itself and the links to it. A mention in an AI answer depends on what the model read about the brand, most of which lives on other people's sites. A company can rank first for its own category and still be missing from AI answers because the comparison articles and forums the model retrieves do not mention it.
The two measurements are complementary. Search Console shows whether the brand's pages are being found; AI visibility tracking shows whether the brand is being named. Searcherries puts both in one project, along with the AI referral traffic from Google Analytics 4, so the three can be read together. A one-off report for any page is available from the free AI visibility checker.
