To measure AI visibility, agree a fixed set of questions your buyers actually ask, run them across the main AI assistants every month, and score each answer: were you mentioned, was the description accurate, and was your site cited? Track the scores over time. Treat the first results as provisional until the question set is agreed.
Key takeaways
- AI visibility is how often, and how accurately, AI assistants mention and cite your company when buyers ask relevant questions.
- Measure with a fixed question set, run monthly across several assistants, scored on mention, accuracy and citation.
- According to TechCrunch, ChatGPT reached 900 million weekly active users in February 2026.
- Researchers found AI search engines lean heavily on third-party sources over brand-owned content, so your own site is only part of the picture.
- Record whether your robots.txt lets AI crawlers in, but keep it out of the score.
What is AI visibility?
AI visibility is how your company appears in answers from AI assistants such as ChatGPT, Gemini, Claude, Perplexity and Google's AI Overviews. It covers three things: whether you are mentioned at all, whether what the assistant says about you is accurate, and whether it links to or cites your site as a source.
It differs from search ranking. A search results page lists ten links and lets the buyer choose. An AI answer usually names a few options and explains them, so being left out means not being considered at all. Being described wrongly can be worse than being left out.
The guide to generative engine optimisation covers how to improve it. This article covers how to measure it.
Why should you measure AI visibility?
Because a growing share of buyer research now happens inside AI answers, before anyone visits your website. If assistants recommend competitors for the questions your buyers ask, you lose deals you never knew existed. Measuring shows where you stand, whether your work is changing anything, and what assistants get wrong about you.
The audience is large. According to TechCrunch, ChatGPT reached 900 million weekly active users in February 2026, up from 800 million in October 2025. Google's own results are changing too.
According to Pew Research Center, Google users clicked a traditional result in 8% of visits when an AI Overview appeared, against 15% without one. Only 1% clicked a link inside the summary. Fewer clicks means traffic reports alone undercount how often buyers meet your brand.
How do you choose the questions?
Choose 20 to 50 questions your buyers really ask at each stage: broad problem questions, category questions such as "best tools for", comparisons with named competitors, and direct questions about your company. Take them from sales calls, customer research and search data, not from guesses. Then agree the set before scoring.
A balanced set covers four types:
| Question type | Example |
|---|---|
| Problem | How do I stop AI making up facts about our product? |
| Category | What are the best options for a company knowledge base? |
| Comparison | How does our company compare with a named competitor? |
| Brand | What does our company do, and who is it for? |
Agreement matters more than it sounds. If the question set changes every month, the scores cannot be compared. Until the people who own the result have approved the questions, label every score as provisional.
How do you score AI answers?
Score each answer on three things. Mention: does the answer name you? Accuracy: is what it says about you correct and current? Citation: does it link to your site or a page you control? Run each question in each assistant, record the full answer, and score it the same way every month.
A simple scale keeps scoring consistent. Give each answer a point for a mention, a point for accurate description and a point for a citation, then total by assistant and by question type. Keep the full text of every answer too, because wrong descriptions are often the most useful finding.
Where the citations come from matters. According to researchers at the University of Toronto, AI search engines show a "systematic and overwhelming bias towards earned media", meaning third-party sources, over brand-owned content. That is why reviews, press coverage and partner pages often decide whether you appear.
How often should you measure AI visibility?
Monthly is enough for most B2B companies. AI answers vary from one run to the next, so a single check means little, while a monthly run of the same questions shows real trends. Run each question more than once per cycle if answers swing widely, and note any model updates that could explain sudden changes.
Answers also differ by phrasing and by assistant. The same question can produce a mention in one tool and nothing in another. Track each assistant separately rather than blending them into one number too early.
Check the technical basics each cycle as well, but record them separately. According to OpenAI's crawler documentation, sites should allow OAI-SearchBot in robots.txt to appear in ChatGPT's search results. A blocked crawler explains a whole platform's poor results, so it is a headline finding. It is still an input, not part of the score.
What do you do with the results?
Use the results to decide what to fix. Wrong descriptions point to missing or outdated facts on your site and elsewhere. Missing mentions on category questions point to gaps in third-party coverage. Missing citations point to pages that do not answer the question directly. Fix the biggest gap, then measure again next month.
Common actions after a first measurement:
- Correct the facts on your own site, in plain, quotable sentences.
- Earn third-party coverage, such as reviews, guest articles and directory listings.
- Write answer-first pages for the questions where competitors are cited and you are not.
- Check robots.txt and consider a llms.txt file, sized honestly.
Skein runs monthly AI visibility measurement on an agreed question set, with scores marked provisional until the client approves the questions; the outcomes page shows what is tracked. That suits a B2B company whose buyers research before they call. A business that wins most work through referrals may not need monthly tracking.
Frequently asked questions
What is AI visibility?
AI visibility is how your company appears in answers from AI assistants such as ChatGPT, Gemini, Claude, Perplexity and Google's AI Overviews. It covers whether you are mentioned, whether the description is accurate and current, and whether the assistant cites your website as a source.
How do you track mentions in ChatGPT?
Agree a fixed set of buyer questions, run them in ChatGPT and other assistants on a regular schedule, and record each full answer. Score whether you were mentioned, described accurately and cited. Repeat monthly with the same questions so the results can be compared over time.
How many questions should an AI visibility test use?
Twenty to fifty questions covers most B2B companies. Include problem questions, category questions, comparisons with named competitors and questions about your own brand. Draw them from sales calls, customer research and search data, and keep the set fixed so scores can be compared month to month.
Why do AI answers change every time?
AI assistants generate each answer fresh, so wording, sources and even recommendations vary between runs and between phrasings. Model updates also change behaviour. That is why one check proves little. Run the same questions on a schedule, and run each more than once if answers swing widely.
Does robots.txt affect AI visibility?
Yes. If robots.txt blocks an assistant's search crawler, that assistant may not be able to find or cite your pages. OpenAI, for example, recommends allowing OAI-SearchBot to appear in ChatGPT search results. Check it every cycle and record it, but keep it separate from the visibility score.
Related reading
- TechCrunch. "ChatGPT reaches 900M weekly active users." TechCrunch, 27 February 2026. https://techcrunch.com/2026/02/27/chatgpt-reaches-900m-weekly-active-users
- Pew Research Center. "Google users are less likely to click on links when an AI summary appears in the results." Pew Research Center, 22 July 2025. https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/
- Chen, Mahe, Xiaoxuan Wang, Kaiwen Chen and Nick Koudas. "Generative Engine Optimization: How to Dominate AI Search." arXiv 2509.08919, 10 September 2025. https://arxiv.org/abs/2509.08919
- OpenAI. "Overview of OpenAI crawlers." OpenAI, accessed 30 October 2026. https://developers.openai.com/api/docs/bots