25 AI-search questions and 6 technical checks against your page, in about 30 seconds.No inflated scores, no vague guesses.
No signup required. No credit card.
Four steps, in the order they run. About 30 seconds, no account. See it in detail →
Paste any URL. We fetch it the way a crawler would, keep the main content, and work out two things separately: what kind of site this is, and what this particular page is about.
The questions people actually put to AI assistants about your topic — spread across cost, comparisons, how-to and trust. Generated from the topic, never from your page: a test written from your own content is one you always pass. Frozen per URL, so re-audits compare.
Each question comes back Covered, Partial or Missing against your actual content. Then 6 technical checks on the page itself. Crawler access is the one we fetch separately, so when your robots.txt can’t be read we say unable to verify and drop it from the average rather than counting it as a pass.
Five of your questions go to Perplexity — and OpenAI on paid plans — and we report which answers cited you. One run, so a snapshot rather than a rate. Every gap comes back with paste-ready content in your page’s language.
You see which questions we tested and how your page scored on each. And when the one check we run off your page — crawler access — can't complete, you get “unable to verify” and the reason rather than a passing grade. Inflated scores and false precision are the standing criticism of this category; here is what we do instead, all of it visible in any report.
When a site blocks our robots.txt request, or the file times out, we report unable to verify and tell you why — “robots.txt returned 403”, “the request timed out” — and drop crawler access from the average. We don’t quietly score it as a pass. It is the only signal that can end up unverified, because it is the only one we fetch separately from your page; the rest are read from HTML we already have. A tool that turns a failed check into a green tick isn’t measuring your site. It’s measuring its own optimism.
llms.txt sits in a separate section called Agent Readiness, worth zero points, with a line saying no major AI provider has confirmed it affects citation ranking. It’s cheap to add and might matter later. That isn’t the same as a ranking factor, and we won’t sell it as one. Your score is 25 questions and 6 technical checks. Nothing else is smuggled in.
Every benchmark shows its sample size. A category median arrives with “18 sites” next to it and a note that these are sites audited with AlpineRank — not a survey of the industry, and people who run audits are people trying to improve. Under five audits of your own, you get a labelled reference point instead of an average invented from three data points.
We put five of your 25 questions to real engines — Perplexity on every plan, OpenAI as well on paid plans — and report which answers cited your domain, on the same row as the coverage verdict. That’s a snapshot. Engines vary between runs, and 0 out of 5 does not mean “never”. Paid plans can re-check any question on demand — another observation, not a rate. We don’t turn a handful of probes into a percentage.
Every date, price and percentage in a generated fix is checked against your page. Anything the page doesn’t support becomes [INSERT PRICE], with a note telling you to fill it in. We’d rather hand you a blank than a confident-sounding number we made up — which is exactly what happened once, and why the check exists.
Read the full methodology, including what this method can't do →
Free, no account, about 30 seconds. You will see the same report shown here — including any check we could not complete.
Beyond the four steps. Where something needs a paid plan, it says so.
Everything you need to know about GEO audits.
An account saves your audits, tracks your score over time, and unlocks the fixes for every gap.
Create free accountNo credit card required