We ask real AI search engines to complete real tasks — book a table, buy a product, get a quote — for real businesses, and honestly count what happens. Nobody else publishes this.
Based on 3 businesses tested since 19 Aug 2026, across 24 scored checks (8 more excluded — see Methodology).
This excludes 8 checks that failed for a technical reason, out of 32 total — never counted as a pass or a fail. Numbers shown as of 20 Sept 2026. Want your own business tested? Run the free audit →
For each business, we send real, natural-language questions — the same kind a customer would type — to live AI search assistants (Perplexity, ChatGPT, Gemini, Claude), the same engines a customer is actually likely to use. We read back exactly what each assistant said and score it against four steps: did it find the business, understand what it offers, find a price or availability, and could a real customer have actually finished the job from that answer.
A check that failed for a technical reason — the assistant wasn't switched on, or the request errored — is never counted as a pass or a fail. It is excluded and the exclusion is stated: this page currently excludes 8 checks on that basis, out of 32 total.
This does not click through a real checkout or booking flow — it is what an AI assistant with web access can find and infer, which is the exact same starting point a real customer asking that assistant would have.