How To Measure Visibility in AI Search

For more than twenty years, measuring search visibility has meant looking at rankings, impressions, clicks and traffic. If your website moved from page three to page one for an important keyword, you could see the improvement. If traffic increased, you could usually trace at least some of it back to search.

AI search is different. A potential customer can ask a detailed question, receive a complete answer and never visit any of the sources used to produce it. Your brand may be recommended, mentioned, cited, compared with a competitor or ignored completely. None of that is properly captured by a traditional ranking report.

This does not mean that AI visibility cannot be measured. It means that we need to measure the answer, not just the link.

What does AI visibility actually mean?

A brand is visible in AI search when it appears in the answer given to a relevant question. That appearance can take several forms.

The brand might be named as a provider. Its product might be included in a shortlist. A page from its website might be cited as a source. Its research, statistics or explanations might be used without the brand being prominent in the main answer. In some cases, the brand may be discussed negatively or described inaccurately.

These are not equal outcomes. A favourable recommendation is more valuable than a passing mention. A citation can establish authority, while an inaccurate description can create a problem despite technically counting as visibility. The useful questions are: Where did the brand appear? How was it presented? Which source supported the answer? How often does the result repeat?

Start with the questions your customers ask

Traditional keyword research is still useful, but people tend to speak to AI systems differently. Searches become longer, more specific and more conversational.

Someone might type “email security software” into Google. The same person could ask an AI assistant, “What is the best email security option for a UK school with a small IT team and a limited budget?” That second question contains a market, a location, an organisation type, a practical constraint and a buying intention.

To measure visibility properly, build a set of questions that reflects the real buying journey. Include broad discovery questions, problem-based questions, comparison questions, questions about cost or suitability, and questions that ask for a recommendation.

For many businesses, an initial set of 30 to 50 carefully chosen questions is more useful than hundreds of loosely related prompts. Keep the wording stable. If you change the question every time, you will not know whether the answer changed because your visibility improved or because you asked something different.

Test across more than one AI platform

There is no single AI search result. ChatGPT, Google’s AI answers, Gemini, Perplexity and Copilot do not necessarily use the same sources or produce the same recommendations. Even two answers from the same platform may differ.

This makes measurement less tidy than checking a conventional ranking. A brand may be strong in one environment and absent from another. One platform may cite the company’s website while another relies on review sites, news articles or a competitor’s comparison page.

Run the same core questions across the platforms that matter to your customers. Record the date, platform, prompt and answer. Where possible, repeat the test over time rather than treating one result as final.

You are looking for a dependable pattern, not trying to prove that a brand always appears.

Measure more than mentions

A useful visibility record should capture several separate signals.

First, record whether the brand appears at all. This gives you a basic mention rate across your chosen questions.

Next, record the type of appearance. Was the brand recommended, included in a list, used as an example, compared with another company or simply cited? A recommendation on a high-intent question carries more weight than a citation attached to a general definition.

Then look at prominence. Being the first named option is not the same as appearing at the end of a list of ten. AI answers do not have rankings in the traditional sense, but position still matters because users are more likely to notice the brands presented early and with more detail.

You should also record sentiment and accuracy. Is the description positive, neutral or negative? Are the products, prices, locations and capabilities correct? Visibility built on old or incorrect information is not a successful result.

Finally, record whether your website is cited. A brand can be mentioned while another source receives the citation. That tells you the AI system knows the brand but does not regard its website as the best evidence for the answer.

Citation visibility deserves its own score

Mentions and citations need to be separated because they tell you different things.

A mention shows that the brand is part of the subject. A citation shows that a particular page has been selected to support the answer. When your site is repeatedly cited for relevant questions, it is evidence that the content is accessible, specific and trusted enough to be used.

Record the exact cited URL, not just the domain. This will show which pages are doing the work. You may discover that one detailed guide earns citations across several questions while your main product pages are rarely used. You may also find that third-party articles are shaping the way AI systems describe your business.

Strong pages can be expanded and kept current. Weak areas can be supported with clearer explanations, evidence, original data, author information and better internal links.

Compare your share of the answer

Your score has little meaning without context. A 20 per cent mention rate might be encouraging in a market where every brand has low visibility. It might be poor if two competitors appear in nearly every answer.

Track the main competing brands against the same prompt set. Count how often each one is mentioned, recommended and cited. This creates a simple view of your share of AI visibility.

The comparison often exposes the real gap. A rival may have one well-researched page that is repeatedly cited. Another may benefit from strong industry coverage. A third may dominate recommendation questions because its positioning is clearer.

Do not copy whatever a competitor has done. Work out what evidence the AI systems appear to be rewarding, then decide how your business can provide something stronger and more useful.

Build a baseline before making changes

Measurement becomes valuable when it shows movement. Before changing the site, run the full prompt set and save the results. This is your baseline.

You can then improve a defined group of pages and test again at sensible intervals. Monthly testing is usually enough for strategic reporting. Weekly checks can be useful while working on a priority topic, although day-to-day movement should not be treated as a trend.

Keep the test conditions consistent. Use the same prompts, platforms and scoring method. Note any major platform change that could affect the results. A sudden shift may reflect a change in the system rather than anything you have done.

The important comparison is not one answer this week against one answer last week. It is the direction of travel across the complete question set.

A practical AI visibility scorecard

There is no universal score that every business must use. The scorecard should reflect commercial priorities.

A straightforward model can award points for a mention, additional points for a recommendation, additional points for appearing prominently, and separate points when the company’s own website is cited. Deductions can be made for inaccurate or negative information.

High-intent questions should carry more weight. Appearing in an answer to “What is GEO?” is useful for an advisory business, but appearing in “Which GEO consultant should a healthcare company use?” is closer to a commercial decision.

The final number matters less than the consistency of the method. Its purpose is to show whether visibility is improving, where the improvement is happening and what still needs attention.

Connect AI visibility with business results

AI visibility should not sit in a separate report with no connection to the rest of the business. Compare it with referral traffic, branded searches, direct traffic, enquiries and sales conversations.

The connection will not always be neat. A person may discover a company in an AI answer, then visit later by typing the brand name or contact the business without clicking a cited link. Exact attribution is difficult, but patterns can still be found.

Ask new prospects how they found you. Monitor whether branded search demand rises in the markets where AI visibility improves. Look at which cited pages receive visits from AI platforms. Check whether enquiries begin using language that appears in AI-generated descriptions of your services.

What should you do with the results?

Measurement should lead to a decision. If the brand is absent from important questions, identify which sources are being used instead. If the brand is mentioned but its website is not cited, strengthen the evidence on the site. If descriptions are inaccurate, make the correct information easier to find and confirm. If a competitor dominates a topic, study the sources behind that advantage.

The point is not to chase every change in every answer. It is to see how AI systems understand the market, where they place your brand within it and whether they trust your content enough to use it.

That is the real value of measuring AI search visibility. It turns a vague question, “Are we showing up in AI?”, into a practical body of evidence. Once you can see the pattern, you can begin to improve it.

About the Author

Simon Royle is an author and Generative Engine Optimisation specialist. He is the author of The GEO Success Series, a practical collection covering how to build an AI-ready website, establish the trust signals AI systems look for and measure visibility in AI search.

Simon works with businesses to understand how they appear in AI-generated answers and what they can do to improve their visibility, authority and chances of being cited. He is also closely involved in the development and marketing of GEOAnalyzer Pro, a platform created to audit websites for GEO readiness and track brand visibility across AI search.

You can find out more at simon-royle.com and geoanalyzerpro.com.