AEO, GEO & AI Search
How to Measure AI Search Visibility (and the Limits of the Data)
A realistic way to measure how people find you through AI-powered search: what Search Console and analytics can show, how to run manual checks, what third-party tools can and cannot tell you, and how to report results honestly.
Last reviewed: 14 October 2026
Business owners and SEO learners now ask a new question: “Are we showing up in AI answers?” It is a fair question, but the data available is limited and sometimes misleading. This guide explains what you can measure reliably, what you can only estimate, and how to report AI search visibility honestly.
Start with what you can measure reliably
Google Search Console
For Google Search, your main source of truth is Google Search Console. Google states that traffic from its AI features, AI Overviews and AI Mode, is included in your overall search traffic and reported in the Performance report under the Web search type. Google’s 2026 guide for generative AI search also refers site owners to a generative AI performance view in Search Console.
Useful views:
- Clicks and impressions over time for your important pages.
- Queries that bring visitors, including longer, conversational queries.
- Pages that gain or lose impressions after changes in search features.
If you have not set up Search Console, follow our free lesson on adding your website to Google Search Console.
Google Analytics 4 (GA4)
GA4 shows what visitors do after they arrive. Visits from AI assistants and AI search tools that link to your site often appear as referral traffic from the tool’s domain, for example chatgpt.com or perplexity.ai, although some visits may arrive without referral information. To review them:
- Open Reports → Acquisition → Traffic acquisition.
- Change the dimension to Session source / medium.
- Look for sources from AI tools and note the landing pages they lead to.
- Compare their engagement and key events (such as enquiry form submissions) with other channels.
GA4 cannot tell you how often you were mentioned in an AI answer without a click. See Google Search Console vs Google Analytics 4 for how the two tools differ.
Enquiries and leads
For a business, the most meaningful measure is still enquiries: calls, WhatsApp messages, form submissions and walk-ins. Ask new customers how they found you and record the answer. If several mention an AI assistant, that is useful evidence even if analytics data is incomplete.
What you can only estimate
Manual checks
You can search your important questions in AI tools and note whether your brand or pages are mentioned or linked. Do this carefully:
- Use the same list of questions each time, written the way real customers would ask them.
- Record the date, tool, location and whether you were logged in, because answers vary.
- Save screenshots as evidence.
- Treat the results as samples, not rankings. The same question can produce different answers for different people or at different times.
Third-party “AI visibility” tools
Some tools run large numbers of prompts and report how often brands are mentioned. They can show trends, but remember:
- They see only the prompts they choose to test.
- They cannot see personalised answers or private conversations.
- Google states that no third-party tool has access to its internal ranking or AI systems.
Use such tools for direction, not as precise measurements.
Common reporting mistakes
- Treating one screenshot as proof of consistent visibility.
- Reporting “AI rankings” as if they were stable positions.
- Comparing numbers from different tools that measure different things.
- Promising clients a number of AI citations or leads. No one can guarantee inclusion in AI answers.
- Ignoring the business outcome (enquiries and sales) in favour of vanity metrics.
A simple monthly report template
| Section | What to include |
|---|---|
| Search performance | Search Console clicks, impressions and top pages (note any changes in search features) |
| AI-related referrals | Sessions and key events from AI tool referrals in GA4, with landing pages |
| Manual sample | Same 10–20 questions, tool, date, whether mentioned or linked, screenshots |
| Business results | Enquiries by source, including “found via AI assistant” if customers say so |
| Actions and notes | Content updates, technical fixes and what to test next |
How to improve what you measure
You cannot control AI answers directly, but you can improve the inputs:
- make sure key pages are indexed and can show snippets;
- answer real questions clearly (answer-first writing);
- add original information and keep facts current (entities, E-E-A-T and original information);
- keep your Google Business Profile complete and accurate for local queries.
Common questions
Can I see exactly which AI answers mention my website?
Not completely. Google reports AI feature traffic within Search Console’s search data, and AI tools that link to your site can appear as referrals in analytics, but no tool shows every mention in every answer. Use manual samples and enquiry data alongside.
Why do AI answers about my business differ between people?
Answers can depend on the wording of the question, location, language, conversation history and changes to the systems themselves. That is why consistent sampling and dated screenshots matter.
What should I report to a client or manager?
Report what you can verify (search clicks and impressions, referral sessions, key events and enquiries) and present manual AI checks as samples. Avoid presenting AI visibility as a stable ranking.
Next step
Measurement works best alongside clear goals. Read SEO, AEO and GEO explained for context, and our content audit guide to decide which pages to improve first.
Sources (checked 14 October 2026): Google Search Central, AI features and your website, Optimizing for generative AI search and Latest documentation updates (AI Mode data in Search Console totals, June 2025).