AI search
five free ways to check your AI search visibility
You can measure your AI search visibility today, for nothing, with five checks. Together they answer the questions any paid dashboard answers, at the scale a small business actually needs, and they double as the baseline you should set before changing anything.
one: the assistant test
Write down the ten questions your buyers actually ask, phrased as they would say them. Put each to ChatGPT, Claude, Perplexity and Google's AI results in fresh sessions, monthly, and log who gets named and which sources get cited. Twenty minutes a month, and it is the single most direct measurement that exists. Trend beats snapshot: absent, mentioned, recommended is the ladder you are climbing.
two: the entity test
Ask the assistants who your company is and what it does. You are checking for three failure modes: not knowing you, confusing you with a same-named business, or describing you outdatedly. Whichever you find is your work order, and it is usually corroboration work, not website work.
three: Search Console's AI reporting
Google has been rolling AI-surface performance reporting into Search Console, as covered in industry reporting through mid 2026, alongside the classic data. Free, first-party, and the nearest thing to an official scoreboard for Google's answer surfaces. If you have not verified your site there, that is step zero for everything, not just this.
four: referrals in your analytics
Visits from chatgpt.com, perplexity.ai and friends appear in your analytics like any referral. Segment them once and watch quarterly. Volumes look small; treat them as a quality signal, because these visitors arrive pre-briefed by an answer that named you.
five: branded search movement
Being named in AI answers often converts into people searching your name directly afterwards, so your branded query volume in Search Console is a lagging echo of answer visibility. Watch it alongside the assistant test; when the first moves, the second follows.
the honest catch
Free measurement is manual, and manual measurement survives on discipline. Answers also vary between runs, so read trends over quarters rather than reacting to any single session. Paid tools automate exactly these checks at scale, and they earn their fee when you are tracking many questions across many markets. Until then, the notebook version tells you everything a decision needs.
Common questions
How many questions should the monthly test cover?
- Ten with real buying intent beats fifty of curiosity. Add questions only when a new one maps to revenue.
What counts as being visible?
- Named in the answer, cited as a source, or recommended outright, in ascending order of value. Log which one you got, not just whether.
When is a paid tool worth it?
- When the manual protocol is genuinely running and its scale hurts: many questions, competitors tracked, several platforms. Buying automation before discipline automates nothing.

Tom Claydon
Co-founder of Nudge
Ask Tom anything about search, he answers on WhatsApp.