Measuring AI Search Visibility

Presence in answers.

The most direct measure: ask the engines the questions your prospects ask, and record whether your business appears. Run a consistent set of prompts — "best [service] in [city]," "who should I call for [problem] near [area]" — across ChatGPT, Perplexity, AI Overviews, and others, on a regular cadence.

It's manual and the answers vary between runs, so the read is a trend, not a single snapshot: are you appearing more often, across more prompts, on more engines than you were a month ago? That movement is the clearest evidence the work is landing.

Share of voice against competitors.

Presence is more meaningful relative to your market. For the same prompt set, track who gets named — you or your competitors, and how often. If three competitors are mentioned every time and you never are, that's the gap. As the work lands, your share of those mentions should climb.

This mirrors how we think about the map pack: it's not whether you exist, it's whether you're in the shortlist the customer actually sees. Share of voice turns a vague "are we visible" into a competitive scoreboard you can move. The three local search surfaces →

Citation footprint growth.

Because AI recommendations rest on corroboration, the growth of your mentions and citations across the web is a leading indicator. More consistent listings, more local press, more references that agree on who you are — these tend to precede improved presence in answers.

Tracking citation footprint is more concrete than tracking the answers themselves, which makes it a useful early signal: if the footprint is growing consistently, presence usually follows, even before it shows up in the prompts.

Referral traffic and the limits.

Some AI engines pass referral traffic you can see in analytics — visits originating from ChatGPT, Perplexity, and AI Overviews. Where it's visible, it's a real, countable signal of impact. But coverage is incomplete: a recommendation a prospect acts on by calling you directly leaves no referral trail at all.

So referral traffic is a floor, not the full picture — it confirms impact when present but undercounts it. This is why we triangulate all four signals rather than leaning on one. Honest measurement here means naming the limits, not papering over them with a fake precision metric.

Track the trend across four signals — and distrust anyone selling a single score.

Set a fixed prompt set, run it on a cadence, and watch presence, share of voice, citation growth, and referral traffic move together over time. No one of them is exact; together they tell the truth. If a vendor hands you a clean "AI visibility score" with no methodology, treat it the way you'd treat a guaranteed ranking — a number designed to reassure, not to inform.