AI Visibility: The Metric That Now Decides Who Gets Found

AI visibility is how often AI answer engines such as ChatGPT, Perplexity, Gemini and Google’s AI Overviews mention, cite or recommend your brand when someone asks a question you should own. It is not a ranking position. It is whether you appear inside the answer at all, whether the facts are right, and who appears instead.

Quick links

• What does AI visibility actually mean?
• Why does rank tracking miss it?
• What does AI visibility look like on each engine?
• How do you start measuring AI visibility?
• What moves AI visibility?
• What should you not read into the numbers?
• FAQ

What does AI visibility actually mean?

It means being present, accurate and recommended in the answers AI engines give about your category. Being indexed is the entry ticket. Being named is the prize.

For twenty years, “visible” meant “ranked”. You typed a query, you got a list, and your position in that list was the score. Everything from reporting to agency retainers was built around it.

An answer engine does something different. It reads a lot of material, writes one response, and mentions a few brands inside it. There is no page two. So visibility now has at least four parts:

• Presence. Are you mentioned at all for the questions your buyers ask?
• Citation. When you are mentioned, is your site linked as the source, or is the engine paraphrasing someone else’s description of you?
• Accuracy. Are the details right: what you do, where, for whom, at what sort of price?
• Share of voice. Out of the brands named for a given question, how often is it you?

Most companies we test are strong on none of those, which is less alarming than it sounds. It means the field is open. Almost nobody is measuring it yet, so a small amount of attention goes a long way.

Why does rank tracking miss it?

Because a rank tracker reports where a link sits in a list, and an AI answer has no list to sit in. You can hold position one for a keyword and be absent from the answer sitting above it.

The scale of that answer box is no longer niche. Google’s CEO said on the company’s July 2025 earnings call that AI Overviews have 2 billion monthly users, with AI Mode past 100 million monthly active users in the US and India.

Then there is what people do next. A Pew Research Center study of US browsing data found users clicked a traditional result in 8% of searches that showed an AI summary, against 15% of searches without one. Clicks on the links inside the summary itself were rarer still, at 1% of visits.

Read that carefully, because it is easy to overstate. It does not say search is dying. It says that when an answer is shown, fewer people go anywhere, so being inside the answer matters more than being underneath it.

Rank tracking also misses everything that happens off Google. A buyer who asks ChatGPT for a shortlist of suppliers never produces a search impression. No dashboard you currently own will tell you that happened.

What does AI visibility look like on each engine?

It differs by engine, because each one gathers and presents sources in its own way. The practical consequence is that you cannot test one and assume the rest.

Google AI Overviews and AI Mode

These sit on top of Google’s index. Google’s own guidance says a page must be indexed and eligible to be shown with a snippet to appear as a supporting link, and that no special optimisations or extra markup are required. Traffic from these features is counted inside your normal web results in Search Console, which is useful, though it is not a separate report. Our post on Google AI Mode and SEO goes further.

ChatGPT

ChatGPT blends what it learned in training with live web retrieval when it searches. Mentions tend to follow how consistently a brand is described across many sources, not just on its own site. There is more in our post on ChatGPT SEO.

Perplexity

Perplexity shows its sources prominently, which makes citations easy to see and count. It is the most transparent of the lot for testing, and a good place to start. See Perplexity SEO.

Gemini and the rest

Gemini draws on Google’s systems but behaves differently from AI Overviews in practice. Claude, Copilot and others add to the picture. You do not need to test every engine monthly. You do need to test the two or three your buyers actually use.

How do you start measuring AI visibility?

Build a fixed set of buyer prompts, run them on a schedule, and log what comes back. You can do the first version by hand in an afternoon.

1. Write 20 to 30 prompts. Phrase them as a buyer would, not as a marketer would. Mix types: category questions (“best [service] for [type of company]”), comparison questions, problem questions, and a few about your own brand name.
2. Run them across your chosen engines. Use ChatGPT, Perplexity, Gemini and a standard Google search. Use a fresh session each time so earlier answers do not colour later ones.
3. Log four things per answer. Were you mentioned? Were you cited with a link? Was the detail accurate? Which competitors appeared?
4. Turn it into a few simple numbers. Mention rate (prompts where you appear, out of all prompts), citation rate, and a short list of factual errors to fix.
5. Repeat monthly, same prompts. The trend is the useful bit. A single run is a snapshot, and a noisy one.

If you want the fuller version of this process, with the technical checks that sit alongside it, our AI search audit checklist walks through it step by step.

Dedicated tracking tools exist and some are decent. They automate the prompt-running and charting. They do not remove the need to decide which prompts matter, which is the part that takes judgement. Start by hand, learn what you are looking for, then decide whether a tool earns its fee.

What moves AI visibility?

Mostly the same things that make a site worth trusting: clear answers, readable pages, and a brand that is recognised across the web. There is no trick, and anyone selling one should be asked for evidence.

Answer-first content. Pages that state the answer in the opening lines, under a heading that matches the question, are easier to lift. This is the cheapest change on the list and often the most effective.

Pages the crawlers can read. If your main content only appears after JavaScript runs, some AI crawlers may never see it. Server-rendered or static HTML sidesteps the problem. This is one reason we build on Astro, and it is also why default WordPress is rarely the right call for a site that needs to be cited.

Entity clarity. Engines repeat brands they can place. Consistent naming, a complete Organization profile, accurate listings and third-party mentions all help. Our guide to entity SEO covers it properly.

Third-party mentions. What credible sites say about you carries weight that your own homepage never will. That makes digital PR and genuine reviews part of visibility, not an optional extra.

Freshness and sourcing. Dated pages that cite primary sources look more trustworthy, to people and to machines.

For the full programme, AI search optimisation is the place to go, and what is AEO covers the underlying idea. If you want the technical side of how language models pick sources, see LLM SEO.

What should you not read into the numbers?

That they are exact. AI answers vary from one run to the next, differ by location and account, and change as models update. Treat any visibility score as a direction of travel.

A few cautions worth holding on to:

• One prompt proves nothing. Mention rate across 25 prompts means something. A single screenshot of a good result means very little.
• Scores from different tools do not compare. Each tool picks its own prompts and its own method. Stick with one set of measures and compare it with itself.
• Mentions are not revenue. Visibility is a leading indicator. Watch branded search, direct traffic and enquiries that mention an AI assistant alongside it.
• Be wary of guarantees. Nobody controls what an AI engine writes. We do not promise citations, and we would be suspicious of anyone who does.

None of this makes the measurement pointless. It makes it a habit rather than a report. Done monthly with a fixed prompt set, it will tell you within a quarter whether the work is landing, which is more than most marketing channels manage.

We run this exact exercise for clients before touching anything else. The baseline is usually the most useful page of the whole engagement, mostly because it replaces an argument about opinions with a table of results. If you are curious about how we approach a rebuild when the audit says the site is the problem, the Hoc Paid Media case study shows a site rebuilt for speed and structure.