Most businesses still can't answer a basic question: does ChatGPT recommend us when a prospect asks the kind of question we'd want to be found for? Rankings and organic traffic don't answer it. Measuring AI visibility properly means tracking three distinct things — share of answer, citation behaviour, and AI-referred traffic — each of which needs a different method.
Share of answer: the AI-era share of voice
Build a fixed set of realistic prompts a prospective customer would actually type — not your brand name, but the underlying question ("best GEO agency for South African B2B companies," "how to reduce zero-click search impact"). Run that same prompt set across ChatGPT, Perplexity, Gemini and Copilot on a recurring schedule, and log whether you're mentioned, where in the answer, and in what tone.
This has to be repeated, not a one-off check. Our own data shows only about 30% of brands mentioned in one AI answer get mentioned again in the very next similar answer, and just 20% stay visible across five answers in a row. A single snapshot will systematically overstate or understate your actual visibility — see the full breakdown on our GEO & AI Search Statistics page.
Citation frequency and position
Beyond "were we mentioned," track where in the source content you were quoted from and how consistently. Citation behaviour skews heavily toward the first third of a page's content, and toward content refreshed within the last two to three months. If your citations are thin, the fix is usually structural — see our companion piece on structuring content for AI answer engines — rather than a measurement problem.
Also worth tracking: platform variance. Citation rates between the AI platform that mentions a brand most and the one that mentions it least can differ by a factor of hundreds. Checking only ChatGPT, however tempting given its traffic share, will miss real movement happening elsewhere.
AI-referred traffic in GA4
Most AI platforms don't cleanly tag referral traffic by default, so it tends to disappear into a generic "referral" bucket unless you build for it deliberately:
Create a custom channel grouping or saved segment filtering on source domains:
chatgpt.com,perplexity.ai,gemini.google.com,copilot.microsoft.com, and any others relevant to your audience.Report on this segment separately, even while its volume is small — our tracked data puts AI referral traffic at roughly 1% of average site traffic, growing by about a percentage point a month.
Compare conversion rate and engagement metrics for this segment against your organic average. The gap is usually large enough to justify the tracking effort on its own.
Putting the three together
Share of answer tells you whether you're in the conversation at all. Citation frequency and position tell you whether your content is doing the specific things that earn a quote. AI-referred traffic tells you what happens once someone does click through. None of the three substitutes for the others — a brand can have decent AI-referred traffic and still be losing share of answer to a competitor gaining ground, or vice versa. Tracking all three monthly, alongside a real GEO strategy, is what turns "we think AI knows about us" into an actual answer.
See our practical AEO/GEO checklist for how this measurement work fits into a broader monthly cadence.