How to Measure AI Search Visibility

To measure AI search visibility, define a fixed prompt set of 20 to 40 buyer questions, run them monthly across ChatGPT, Perplexity, Gemini and Google AI Overviews, and record three metrics: mention rate (how often your brand is named), citation share (which of your pages get cited as sources), and competitive share-of-voice (your mentions versus named competitors). Repeat monthly to track trend.

Why traditional SEO tools cannot measure this

SEMrush, Ahrefs and Search Console measure Google blue-link positions. None of them measure whether ChatGPT names your business when a buyer asks for recommendations. That is a separate, increasingly important question that needs its own instrumentation.

The good news is that the measurement method is straightforward and can be run by anyone with a methodical approach.

The three core metrics

  • Mention rate: of N prompts, in how many does your brand get named? Reported as a percentage and as a count.
  • Citation share: when your brand is named, which page is cited as the source? This shows which of your pages the engines trust most.
  • Competitive share-of-voice: of total brand mentions across your prompt set, what percentage are yours versus each named competitor?

How to build a useful prompt set

  • Start with 20 to 40 prompts. Smaller misses signal; larger gets noisy.
  • Cover three categories: direct buyer prompts ('best AI consultant for UK insurance'), problem-led prompts ('how do I reduce motor claims cycle time'), and comparison prompts ('Optimus Consulting vs [competitor]').
  • Use the actual phrasing buyers use, not internal jargon. If your team calls it 'claims orchestration' and buyers call it 'claims automation', use the latter.
  • Lock the prompt set. The same prompts every month is what gives you trend; changing prompts gives you noise.

Which engines to test

  • ChatGPT (test on both GPT-4o and GPT-5 if available; results differ)
  • Perplexity (strongest signal of live retrieval; cites sources transparently)
  • Google Gemini (often surfaces in Google AI Overviews)
  • Google AI Overviews directly (test in incognito with a UK IP)
  • Anthropic Claude (growing share, especially in B2B)

Monthly cadence and what to look for

Run the prompt set on the same day every month. Record results in a spreadsheet with columns for engine, prompt, named brands and cited sources. After three months you have a trend; after six you have a story.

Look for: mention rate climbing or falling, new competitors appearing in answers, your citation sources shifting (a sign of which pages are gaining authority), and prompts where you are never named (the gaps to fix first).

What a working measurement system looks like

A UK insurance broker we work with runs a 28-prompt set on the first Monday of every month across four engines. Total time: about two hours including data entry. Outputs: a single dashboard showing mention rate (up from 4% to 38% over six months), citation share (their service pages cited in 71% of mentions), and share-of-voice versus three named competitors (from fourth to second). That dashboard now drives the GEO content calendar: prompts with zero mentions become the next month's content priorities.

Frequently Asked Questions

Do I need a paid tool to measure AI visibility?

No. A spreadsheet and two hours per month is enough to start. Paid tools (Profound, AthenaHQ, Otterly) speed it up and add competitive features but are not required to begin.

How long before I see a trend?

Three months minimum. AI engines retrain and re-rank on cycles of weeks; a single month is too short to separate signal from noise.

Should I include negative prompts?

Yes, sparingly. One or two prompts like 'problems with [your category]' help you see whether you are mentioned in critical contexts, and whether competitors are weaponising any weak spots.

What counts as a mention?

Any time the engine names your brand in the answer, including in lists, recommendations, comparisons or examples. Citations of your website without naming the brand also count, recorded separately.

Can I automate this?

Partially. The API calls to OpenAI, Anthropic and Perplexity can be automated. Google AI Overviews still need manual checks because they vary by IP, account and session. We typically blend automated runs with monthly manual spot-checks.

Loading the full page. If it doesn't load, JavaScript is required.