Your first AI visibility audit
A practical walkthrough: from one URL to a clear picture of where you stand in AI answers.
Step 1 — Pick the questions that matter
Start with how buyers actually ask. List 15–30 real questions in your category: the "best tool for…" discovery prompts, the "alternatives to [competitor]" comparisons, and the high-intent "which X is best for Y" questions. These are the answers where being named changes revenue.
Step 2 — Read the answers, across engines
Run those questions across the major engines — the same question can produce a different winner on each one. For every answer, note three things: are you mentioned, are you recommended or just listed, and who gets named instead.
Step 3 — Trace the sources
For the answers where a competitor wins, look at what the engine cited. A pattern usually appears fast: a couple of review sites or comparison pages doing the heavy lifting. Those are your citation targets.
Step 4 — Turn it into three moves
Don't try to fix everything. From the audit, pick the three highest-impact moves — usually some mix of a comparison page worth publishing, a citation worth earning, and a technical gap worth closing. Ship those, then re-run and measure the change.
How to write a question worth tracking
The questions that matter are the ones a buyer would actually type, not the ones you wish they would. "Best project management tool for a small agency" is a real question. "Leading enterprise workflow orchestration platform" is a positioning statement wearing a question mark, and nobody asks it. Three tests: would a stranger type this, does the answer plausibly name several brands, and would being named in it change whether someone buys? If any answer is no, drop it — a tracked question that nobody asks costs you the same as one that matters.
How many, and how often
Fifteen to thirty questions is enough to see a pattern in most categories, and the reason is arithmetic rather than taste. Each asking is one observation, and a mention rate built from a handful of them has a margin so wide it cannot distinguish good news from bad. Ask the same set repeatedly rather than expanding it endlessly: thirty questions asked every day tells you far more than three hundred asked once, because only the first gives you a rate you can trust and a baseline to compare against.
What to write down
For each answer record four things and you will have everything the rest of the work needs: were you named at all, were you recommended or merely listed among others, which brands were named instead, and which sources the engine cited. The fourth is the one people skip and the one that turns an observation into an action — a missing mention tells you there is a problem, but the cited source tells you where it lives.
A shortcut
Brand Climb automates steps one through three from a single URL and hands you step four as a ranked plan, free to see your score and with no card. But the method above works perfectly well in a spreadsheet, and the discipline is what matters: same questions, asked repeatedly, with the sources written down.