TapCheck reads your site the way ChatGPT, Perplexity and Google AI Mode do — robots.txt, schema, answer blocks, entity signals — and scores it out of 100 in about ten seconds. No email needed for the score.
Twenty of your buyers’ questions, three runs each, logged out and market-pinned. Your top competitors benchmarked signal by signal. Every fix in order for the next 90 days. Thirty minutes with Tymon to scope it.
Each pillar answers one question an answer engine has to answer before it can cite you. The score covers what can be measured from your site alone; whether AI actually names you is tested separately.
Can live retrieval agents — ChatGPT-User, OAI-SearchBot, PerplexityBot, Claude — open your pages? Training crawlers are scored separately, because blocking them costs far less.
Organization schema that says who you are, page-level FAQ, Article and Product markup, and JSON-LD that actually parses.
Headings followed by a 40–60 word direct answer, question-led headings, lists and tables, and enough pages for a model to cite.
sameAs links to your profiles, an About page, named people and authors, contact details, a plain description of the company.
Totals renormalise to 100. Anything we couldn’t fetch is marked “not tested” and left out, never scored zero. Full protocol: methodology.
TapCheck measures technical readiness to be cited by AI answer engines across four pillars: AI crawler access (robots.txt for live retrieval agents and training crawlers, llms.txt, content in the HTML), structured data (Organization and page-level schema), answer extractability (headings followed by direct, quotable answers, lists, tables and citable surface), and entity and authority (sameAs links, About page, named people, contact details). The score is out of 100.
No. TapCheck scores what a machine can read on your site; it does not run questions through AI engines. Whether AI actually names you is measured separately. The free emailed report adds 20 of your buyers' questions tested on Google AI Mode, and the paid audit runs twenty questions three times each across engines.
GPTBot is a training crawler: blocking it slows how a model learns about you over time. ChatGPT-User, OAI-SearchBot and PerplexityBot are retrieval agents that fetch a page while answering a live question: blocking them means the model cannot cite you even when it wants to. TapCheck scores the retrieval block as critical and the training block as minor, and recognises Cloudflare's content-signal policy as a deliberate choice.
80 to 100 is Strong, 60 to 79 Developing, 40 to 59 At Risk and below 40 Critical. Most B2B sites land between 25 and 50, so a low score usually reflects an industry-wide gap rather than an unusual failure.
Yes. The score and the three highest-priority fixes are free and need no sign-up. The full report, with every check, every fix and the buyer-question test, is free too and goes to your email. The full audit with competitor benchmark and a 90-day plan is a paid engagement, scoped on a short call.