How to measure AI Search Visibility (open-source script)

• Define your buyer-intent query list (10-30 queries, written exactly the way buyers phrase them, not keyword-style)
• Pick the engines you care about (start with ChatGPT, Perplexity, Gemini)
• Run each query against each engine, programmatically or manually
• Record the prompt, engine, UTC capture time, available session settings, market and language, answer and cited URLs. Note brand mentions, linked domain citations, source-list position and surrounding context separately. Check cited sources for relevance and support. Calculate citation presence as observed runs with a linked domain citation divided by all observed runs in the same question-and-engine set. Report repeat counts and failed or unavailable runs separately. Keep analytics referrals and attributable revenue separate from answer observations.
• Repeat weekly or monthly to track movement

• Example planning target, not an engine standard: 80+ on the named readiness rubric. Report its version and inspected checks separately from live citations.
• Example planning target: citation presence in 60%+ of the selected priority queries on at least one recorded engine; report query count and repeated-run conditions. This is not a universal success threshold.
• Example planning target: citation presence in 30%+ of the selected priority queries on at least three recorded engines; report query count and repeated-run conditions. This is not a universal success threshold.
• Domain appears as a top-3 source on at least one query per priority cluster

Most teams treat AI search visibility as unknowable, so they do nothing, and doing nothing is how you stay invisible. The moment you put a number on it, the work becomes obvious. A monthly AI search visibility score tells you which questions you own, which a competitor owns, and which are unclaimed, and that ranking is a to-do list in priority order. You stop debating whether AI answers matter and start closing specific gaps.

Repeated comparable observations can show a pattern worth investigating. Record the implementation and later positive, negative or unchanged answers. A movement after one edit is not proof that the edit caused it; use the result with the buyer problem, source evidence and commercial feedback to choose the next action.

The tooling is secondary. You can run the whole method by hand for twenty questions in an afternoon, or script it against the engines for a larger set, but the discipline is identical: define the questions, measure where you stand, ship one improvement, re-measure. Brands that do this quietly pull ahead, because AI search visibility is still uncontested for most questions. The window to own your category in AI answers is open precisely because so few competitors are measuring at all.

The bottom line is that AI search visibility is a channel like any other, and channels reward the people who measure them. Define your questions, put a number on where you stand across ChatGPT, Perplexity and Gemini, and manage that number the way you manage pipeline. The teams that start measuring now will own their categories in AI answers while everyone else is still guessing whether it matters. To compare engines without treating their observations as interchangeable, see citation patterns across Perplexity, ChatGPT and Gemini.

There is no downside to starting small. Pick five questions this week, run them through the three engines, and write down what you find; even that tiny sample usually reveals a gap you can fix in an afternoon. Momentum in AI search visibility comes from the habit, not the scale, so the fastest move is simply to measure something today rather than plan a perfect system for later.

Start today, measure monthly, and let the number, not opinion, decide where your AI search visibility effort goes next; that single habit is what separates the brands that get cited from the ones that keep hoping they will be.