Run 79 deterministic checks and see what ChatGPT, Claude and Perplexity can actually read. Get a scored fix plan with evidence — free, without an account or sales call.
An assistant does not browse like a person. It fetches a few URLs, processes the returned HTML and decides in milliseconds whether your page contains a citable answer. Three causes often mean it does not:
None of these problems appears in a normal rankings report. You notice them through what is missing: your company is simply never mentioned.
Signals that do not apply to your website receive a reasoned N/A status and do not count. This makes the score reflect your website rather than a generic checklist.
Whether ten named AI crawlers can reach you at all — and whether they receive the same page as a regular visitor.
The head tags and heading structure every engine reads before deciding what your page is about.
The machine-readable version of your website. Without it, an assistant has to guess who you are.
Whether there is a passage an assistant can lift as a clear, attributable answer and cite you for.
Whether the website resolves to one real, recognisable organisation an answer engine can name with confidence.
Whether the engine finds substantive content or mainly navigation and interface elements dressed up as a page.
We fetch your homepage, robots.txt, sitemap.xml and llms.txt in parallel from a server — the same way a crawler arrives. A time-boxed second wave then visits the declared sitemap, your About, Contact and FAQ pages, and makes one extra request identifying itself as GPTBot.
No model is asked for an opinion. Every result is a measurement with verifiable evidence. Signals that cannot apply to your website — such as Product schema on a site without products — become N/A with a reason and do not count towards the score.
Related problems are grouped into root causes with their combined point value, an estimated score after the fix and copy-ready snippets for JSON-LD, robots.txt and llms.txt. The full report — every finding and every fix — is sent by email.
What the scan deliberately does not do: run a headless browser, call Lighthouse, ask a language model for an opinion or crawl your entire website. It reads your homepage and up to three important pages — the same shallow pass an answer engine makes.
What the scan measures, how the score works and where its limits are.
The scan shows what can improve. If you would rather have it fixed than merely saved, discuss it directly with the people who also build it.
Schedule a no-obligation callWe build and operate digital products ourselves and run the same scan on our own websites. If you want an experienced team to implement the fix plan, we can discuss where the greatest gains are.
