How we measure GEO. Every number, and where it comes from.
Plenty of agencies sell AI visibility with nothing behind it. This is exactly how we measure it, so you can check our work, or anyone else's.
- We measure GEO with a fixed set of buyer questions, run across the major AI engines on a regular schedule.
- The core metrics are mention rate, citation share, recommendation share and answer accuracy.
- We connect it to the business through AI referral traffic, a “how did you find us?” field and enquiry tracking in your CRM.
- You see your baseline in week one, then a full report every month.
Your share of all the sources cited in AI answers to a fixed set of questions. If AI cites 40 sources across your tracked questions and 10 are yours, your citation share is 25%.
Related: mention rate, recommendation share.
How do you measure the success of GEO?
In five steps, the same way every month, so the numbers compare fairly and you can check every one of them.
Build the question set
50 to 150 questions your buyers ask, in their words, grouped by service, buying stage and market. Agreed with you before we start, then kept fixed so months compare fairly.
Choose the engines
ChatGPT, Claude, Perplexity, Gemini, Microsoft Copilot, Google AI Overviews and AI Mode, weighted by where your buyers actually ask.
Run every question repeatedly
AI answers vary from one run to the next, so we run each question several times per engine and report the pattern, never a single screenshot.
Score every answer
Were you mentioned? Cited with a link? Recommended? Described accurately? Who appeared instead, and which sources did the AI lean on?
Connect it to enquiries
AI referral traffic in your analytics, a “how did you find us?” field that lists AI assistants, and enquiry and pipeline tracking in your CRM.
The GEO metrics we report, and how each is calculated.
Six numbers. Each one has a plain definition and a formula, so there's nothing to take on trust.
AI visibility score
A single 0 to 100 summary of how visible you are across engines, combining mention rate, citation share and accuracy, weighted by where your buyers ask.
Mention rate
The share of tracked answers that mention your firm by name.
Citation share
Your share of all the sources cited across tracked answers.
Recommendation share
How often AI actively recommends you, not just mentions you, compared with competitors.
Answer accuracy
The share of answers that describe your services, locations and people correctly.
AI referrals and enquiries
Visits from AI assistants and the enquiries they produce. ChatGPT, for example, adds utm_source=chatgpt.com to links it shares.
One screenshot proves nothing.
Ask ChatGPT the same question twice and you can get two different answers. That's why we never present a single screenshot as success, and why you should be wary of anyone who does. We report patterns across repeated runs, and link every number to the answers behind it.
What we can't promise, and what we can.
What people ask about measuring GEO.
How do you measure the success of GEO?
By tracking a fixed set of buyer questions across the major AI engines every month, measuring mention rate, citation share, recommendation share and accuracy, then connecting AI referrals and enquiries to the work in your analytics and CRM.
What's a good citation share?
It depends on how crowded your sector is. We benchmark you against the competitors AI names most often, and measure the gap closing month by month.
How do you measure the ROI of GEO?
By comparing the cost of the work with enquiries and revenue from search and AI, using AI referral data, enquiry source fields and your CRM. Some influence happens without a click, so we also watch branded searches and direct enquiries.
Can we see the raw answers?
Yes. Every report links to the underlying answers, so you can read exactly what AI said.
How often do you report?
A baseline in week one, a short update every Friday during the pilot, and a full report every month.
Get your baseline, free.
Your snapshot shows where you stand today, in the same terms we'd report every month. Within 48 hours, reviewed personally.
James KevanCo-founder · Strategy & client relationships