How to measure your brand's visibility in AI search

What AI search visibility means

When someone asks an AI assistant which agency, tool or product to choose, the answer often names a handful of brands and sometimes links to sources. AI search visibility is your presence in those answers: how often you are named, whether your pages are cited as sources, and whether the way you are described is accurate and favourable.

It overlaps with SEO but isn't the same thing. A page can rank well and still be left out of AI answers, and a brand with modest rankings can be mentioned often because other trusted sites talk about it. That is why a separate measurement is worth having.

Why rankings aren't enough any more

Traditional reporting assumes a searcher sees a list of links and clicks one. AI Overviews and assistants change that: the searcher may get a complete answer, with your brand recommended or missing, and never visit a results page at all.

If you only track rankings and organic sessions, you can miss both a gain (you are being recommended, but the click goes elsewhere or nowhere) and a loss (a competitor is now the default suggestion). Measuring AI visibility closes that gap.

Step 1: build a prompt set from real questions

Your prompt set is the list of questions you will put to each assistant every time you measure. It is the most important decision you make, because it defines what visibility means for your business.

• Start with real language: sales call notes, support tickets, Search Console queries and the questions people ask in communities.
• Group prompts by topic (for example, each product line) and by stage: learning ("what is…"), comparing ("best … for …") and choosing ("… vs …", "is … worth it").
• Include unbranded prompts. Asking about your own brand by name tells you how you are described, but unbranded prompts tell you whether you are recommended at all.
• Keep the set stable. Add new prompts as a separate group so your trend line stays comparable.

Step 2: choose the assistants that matter to your audience

Most teams start with Google's AI Overviews and AI Mode, ChatGPT, Gemini, Perplexity and Microsoft Copilot. Prioritise the ones your customers actually use; a B2B software buyer and a local consumer may lean on very different tools.

Note the conditions you test under, such as country, language and whether you are signed in, and keep them the same each time. Location and personalisation can change answers.

Step 3: record the same four things for every answer

For each prompt and assistant, capture the full answer and score it consistently:

• Mention: is your brand named? If so, where in the answer (first recommendation, in a list, or a passing reference)?
• Citation: is one of your pages linked or listed as a source? Which page?
• Description: is what the assistant says about you accurate and current, and is the tone positive, neutral or negative?
• Competitors: which other brands are named, and which sources are cited for them?

Step 4: sample repeatedly and measure the trend

AI answers aren't fixed. The same prompt can produce different brands on different runs, so a single check proves little. Run each prompt several times per session and repeat sessions on a regular schedule, such as weekly or fortnightly.

Then summarise. Share of voice (the percentage of answers in which you are mentioned, compared with competitors) is the headline number most teams report. Track it per topic and per assistant, alongside how often you are cited as a source.

Manual checks or a tool?

A spreadsheet works for a small prompt set and is a good way to learn what the answers look like. It becomes slow and inconsistent once you track dozens of prompts across several assistants.

AI visibility tools automate the runs, store the answers and calculate share of voice. When comparing them, check which assistants they cover, how prompts are chosen, how often each prompt is run, and whether you can see the full answers and cited sources rather than just a score.

How to improve what you find

Measurement shows where you are missing. Improving it looks a lot like building authority in search generally:

• Make the facts about your brand clear and consistent: what you do, who for, where, and how you differ. Use plain language and structured data on your own site.
• Publish content that answers the prompts directly, with a clear summary near the top, so it is easy to quote.
• Earn mentions on the sites assistants cite for your topic, through digital PR, partnerships, reviews and expert commentary.
• Check the sources cited for competitors. They often show exactly which publications and directories carry weight in your market.

Common mistakes to avoid

• Judging visibility from one screenshot of one answer.
• Only testing branded prompts, which hides whether you are recommended for the things customers actually ask.
• Changing the prompt set every month, which breaks the trend.
• Reporting a single score without the answers behind it, so nobody can see what to fix.

Frequently asked questions