AI visibility is useful evidence, but it is not a lead. A sensible GEO measurement plan keeps three questions separate: Did an AI system mention or cite us? Did someone visit our website? Did a suitable prospect enquire? Track those layers together, but never treat one as proof of the next.
What should you measure first: mentions, citations or recommendations?
Start with a repeatable visibility log. Choose a fixed set of customer-style questions about the services you sell. Run the same questions on the AI surfaces relevant to your market, such as Google AI Overviews, ChatGPT, Gemini and Perplexity. For each check, record the date, surface, exact prompt, whether your business was named, whether your website was linked or cited, which page was referenced, and which competitors appeared.
Keep the wording and checking conditions consistent. AI responses can vary, and a single successful answer is not a market share figure. A dated log shows whether your business is appearing more often across your chosen questions; it does not show how many people saw the answer or whether they trusted it enough to contact you.
This is close to what Engage24’s service page calls AI citation analysis: assessing current mentions and citations across Google’s AI Overview and leading AI assistants. Its complimentary check also describes a snapshot based on 20 customer-style prompts and four AI search surfaces, subject to each surface’s availability at the time. Treat that as a time-stamped diagnostic, not a permanent score or guarantee. Engage24’s GEO service
How can Search Console show Google AI visibility without proving a lead?
For Google Search, use Search Console’s Generative AI performance report where it is available. Google says this report helps site owners understand how content performs in generative AI features on Google Search and Discover. 1
Google also explains that AI Overviews and AI Mode are included in the overall Search Console Performance report under the Web search type. That report can show impressions, clicks, queries and pages, but it does not turn an AI mention into a conversion. 2 Compare the same date ranges and pages over time, and annotate meaningful content or technical changes. Do not add clicks from Search Console to AI-assistant mentions from your own prompt checks as if they were one metric.
Google says eligibility to appear in AI features requires a page to be indexed and eligible to appear in Search with a snippet, but meeting requirements does not guarantee crawling, indexing or serving. It also says there are no additional special AI optimisations required beyond foundational SEO best practice. 2 That is why a visibility increase should be reported as an observed change, not as a promised outcome.
How do you measure visits and qualified enquiries separately?
Use your analytics platform for what happens after a visitor arrives. Google distinguishes Search Console’s pre-visit Search activity from Analytics data about page views, sessions and actions on the site. The two systems use different metrics, so their totals will not match exactly. 3
Define a small conversion set before you review results. For an SME, that might include a completed quote form, a booked consultation, a tracked phone call or a clearly qualified email enquiry. Mark the source as accurately as your setup allows, review landing pages and engaged sessions, and check the actual enquiry record rather than counting every contact as a sales opportunity.
Then report a simple funnel for the same period:
- AI visibility: mentions, cited pages and answer observations.
- Visits: Search clicks, analytics sessions and landing-page engagement.
- Qualified enquiries: enquiries meeting your agreed fit and intent criteria.
The numbers will not always move together. AI answers can mention a company without a click. Search traffic can rise while enquiries stay flat if the page, offer or follow-up is weak. Enquiries can arrive through a direct visit after someone first discovered the brand in an AI answer. Record that uncertainty instead of assigning unsupported credit.
What are the limits of a GEO measurement report?
A good report states its test set, dates, surfaces and definitions. It does not claim universal AI visibility from a handful of prompts, infer leads from citations, or promise rankings, inclusion or recommendations. Google warns that third-party tools cannot access its internal ranking or AI systems, so treat vendor scores as workflow aids rather than official Google measurements. 1
If you want a baseline, Engage24’s current GEO route offers an AI readiness audit and a tailored improvement plan, with a free GEO audit or quote request available for the next step. Ask for the tested prompts, surfaces, date and definitions so the result can be compared honestly over time.
References
Featured image: “Two colleagues reviewing analytics on a laptop screen in a contemporary workspace,” by Vitaly Gariev via Pexels; used under the Pexels licence.

