Measuring AI citations when there's no analytics for it

You measure AI citations sideways, because there's no Search Console for ChatGPT: prompt-test by hand each month, track brand mentions as a leading indicator, and watch referral traffic from AI domains. The citation itself happens inside a conversation you never see — Google Analytics and Search Console cover clicks and impressions on the open web, but neither tells you that ChatGPT recommended you to someone last Tuesday. So you measure the things around it. Here's what I actually do.

1. Prompt-test it by hand

The most honest signal is also the most manual: take the questions your customers actually ask, and ask them — in ChatGPT, Perplexity, Google AI Mode. Are you cited? Mentioned? Absent? Note it, with the date and the prompt, and repeat monthly. It's crude and it doesn't scale, but it's real, and it tells you which queries you're winning and which you're invisible for.

A concrete example. For Fingo Private Tours, a tour operator in Hội An, I test twelve prompts monthly: "best private tour guide in Hoi An," "Hoi An day trip from Da Nang," "Vietnam travel itinerary with a local guide," and nine more pulled from actual search queries. In March 2026, Fingo appeared in zero of twelve ChatGPT responses. By June, after three months of GEO work — structured data, answer-first content, local citations — it showed up in four. Not dominant. But present. And the prompts where it appeared were the ones where the underlying content was strongest. That correlation told me more than any score could: the citations followed the content, not the other way around.

2. Track brand mentions, not just links

AI answers are downstream of what the web already says about you. So watch for your brand showing up — in articles, forums, roundups, reviews — because those are the raw material the engines synthesise. Rising mentions are a leading indicator of rising citations.

3. Read the faint referral signals

A trickle of referral traffic now comes from chatgpt.com, perplexity.ai, and friends — small, often under-attributed, but worth a segment in your analytics. Watch for two patterns: those referrers creeping up, and unexplained bumps in direct traffic after you've been visible in an answer (people hear the name, then type it).

4. Watch the inputs, since the output lags

Citations trail authority by months — I've watched that lag firsthand. So the most useful day-to-day metric isn't citations at all; it's the things that cause them: brand mentions, quality links, reviews, topical depth. Move those and the citations follow.

What I'd do differently

When I started measuring AI citations for Sartoro, I made a spreadsheet with twenty prompts and tested them weekly. That was too much, too often. The data was noisy — ChatGPT's answers shift day to day, and testing twenty prompts every Monday produced a chart that looked like a seismograph. Nothing was actionable.

Now I test monthly, not weekly, and I test fewer prompts — the eight that actually map to revenue. "Best tailored suit in Hoi An." "Where to buy linen shirts in Vietnam." The ones where a citation could plausibly become a customer. The rest I check quarterly, out of curiosity, not strategy.

The other thing I'd change: I'd start tracking brand mentions before starting the GEO work, not after. When I began the Fingo measurement, I had no baseline for mentions. I knew citations went from zero to four, but I couldn't prove the mention trajectory that caused it. Now I set up the mention tracker on day one, so the causal chain is visible: mentions rise, then citations follow, usually six to eight weeks later.

The honest framing

It's early, and it's noisy. Treat AI citation as a KPI you sample — a monthly manual sweep plus a mention tracker — not a dashboard you refresh. Anyone selling you a precise "AI visibility score" is mostly selling you a proxy with a confident UI.

If you want a hand standing up that measurement — or the GEO work that earns the citations in the first place — that's what I do.