No Rankings to Check: How to Measure Your AI Visibility Anyway

SEO spoiled us with scoreboards. Rankings, impressions, click-through rates: whatever else was hard about it, at least you could see the numbers move. AI visibility offers no such dashboard. Ask ChatGPT who the best accountant in town is and there's no report telling you where you placed or why.

But unmeasurable and undashboarded aren't the same thing. You can absolutely measure your AI visibility. You just have to collect the evidence yourself, and the method is simple enough to run monthly without tools or budget.

Build a fixed question set

Measurement starts with consistency. Write down five to eight questions and never change them, so month-over-month comparison means something. A good set covers three angles. Direct: "tell me about [your business]." Category: "best [service] in [your area]" and one or two variations a real customer would use. Decision-stage: "is [your business] good for [specific need]" and "[your business] vs [main competitor]."

Run the full set through ChatGPT, Gemini, and Perplexity. Same wording, fresh conversation each time so earlier chat history doesn't color the answers. Twenty minutes, once a month.

Score three things: presence, accuracy, sentiment

For each answer, record three simple judgments. Presence: were you mentioned at all, and were you a headline recommendation or an afterthought? Accuracy: of the facts stated about you, how many were right? Wrong hours, dropped services, and old addresses all count against. Sentiment: is the framing enthusiastic, neutral, or hedged with warnings?

Score each answer simply, even a 0-to-2 scale per dimension works, and total it up per engine. The point isn't statistical rigor. The point is a number that moves when reality moves, so you can tell whether the work you're doing is landing.

Track who else gets named

Your absence is only half the story. When an engine skips you, it recommends someone. Write down who. Over a few months you'll see which competitors own which questions, and often why: they have the recent reviews, or the one comparison article that keeps getting cited, or simply cleaner data.

This is competitive intelligence nobody is hiding. The engines will explain their reasoning if you ask a follow-up: "why did you recommend them?" The answer frequently names the exact sources doing the work, which converts directly into your to-do list.

Watch the traffic that already tells you

Your analytics contain a rough AI-visibility signal today. Referral traffic from chatgpt.com, gemini.google.com, and perplexity.ai shows up in most analytics tools, and its trend line is meaningful even when the raw numbers are small. So is a subtler tell: direct traffic and branded searches rising after an AI tool named you, because plenty of people read an answer, then go type your name themselves.

Add one human-scale metric if you can: ask new customers how they found you, and actually log the answers. "ChatGPT told me about you" is a sentence owners have started hearing, and counting it costs nothing.

Keep the scorecard boring

One spreadsheet. One row per month. Columns for each engine's presence, accuracy, and sentiment scores, the competitors named, and the referral trend. That's the whole system. It takes half an hour a month and after one quarter you'll have something rare: actual evidence about how the fastest-growing discovery channel treats your business.

Most owners are still flying blind here, reacting to a screenshot a friend sent or a bad answer they stumbled into once. A boring little scorecard beats that by a mile. You can't manage what you never look at, and looking, it turns out, was the only hard part.