What does AI visibility mean?
In classic search, visibility can be measured through rankings: a page holds a certain position for a keyword. In AI answers, there is no fixed position. A system writes a new answer every time, names three providers on one occasion and ten on another, and links to different sources. AI visibility is therefore measured as a share: in how many answers to relevant questions does the brand appear, and how is it described?
AI visibility is the central metric for Generative Engine Optimisation (GEO).
The key metrics
- Mention rate: the share of answers in which the brand is named.
- Citation rate: the share of answers in which your own website is linked as a source. A brand can be named without its website being linked, and vice versa.
- Share of voice: your own mentions divided by the mentions of all brands considered in the same prompt set.
- Accuracy: the share of answers in which details about the brand are correct, such as address, prices or services.
- Context: whether the brand is recommended, mentioned neutrally or described critically.
It helps to separate results by prompt type: questions containing the brand name, generic questions about solutions or providers, and comparison questions. High scores on brand questions say little about whether a brand is recommended in response to neutral questions.
Why AI answers vary
At the end of 2025, SparkToro and Gumshoe had around 600 volunteers enter twelve prompts into ChatGPT, Claude and Google’s AI, 2,961 runs in total. The result: for ChatGPT and Google’s AI, the chance of getting the same list of brands twice across a hundred repetitions was below one in a hundred. The same list in the same order appeared roughly once in a thousand runs.
The authors conclude that positions in AI answers are of little use as a metric, whereas a visibility share across dozens to hundreds of prompts, each run several times, is a reasonable one. Individual manual spot checks therefore do not give a reliable picture.
How AI visibility is measured
- Define a prompt set: questions real customers would ask, mixing brand, solution and comparison questions.
- Choose systems and markets: for example ChatGPT, Gemini, Claude, Perplexity as well as AI Overviews and AI Mode, each for the relevant countries and languages.
- Query repeatedly: run the same prompts at fixed intervals and several times to even out fluctuations.
- Analyse: record mentions, source links, competitors and accuracy, and aggregate them into shares.
- Spot-check: tools do not recognise every variant of a name. Manual checks are part of the process.
Google also provides its own data. The Generative AI performance report in Search Console shows impressions in AI Overviews and AI Mode, but no clicks (Search Console Help). Visits from chat applications appear in web analytics as referrals, provided the browser passes on where the visit came from. Our AI visibility tool offers a quick start, and AI monitoring provides ongoing measurement.
An example from a client project
For a winery with an online shop, we analysed 50 prompts across ChatGPT, Claude and Gemini, 150 answers in total. For brand questions, the winery appeared in 21 of 24 answers. For generic questions, such as those about wineries or wines from the region, it appeared in only 26 of 126. A regional competitor appeared in 59 of 126, giving it more than twice the mention rate.
There were errors too: Gemini gave a wrong address and wrong opening hours, and Claude quoted incorrect shipping costs. You can read more about such errors in our article on AI hallucinations.
Improving AI visibility
Measurement shows where the gaps are. If a brand is missing from generic questions, it needs content that answers exactly those questions (AI SEO) and mentions in the sources AI systems rely on. If details are wrong, the work usually starts on your own website: make facts consistent, keep them up to date and publish them in HTML rather than only in PDFs.
Would you like to measure and improve your AI visibility and need support? Then get in touch or read more about our AI monitoring! Would you like to find out more or have your team trained on this topic? Then take a look at our GEO seminar.
