Methodology
No hidden magic and no invented facts. Here is exactly what we read, how it is judged, and what the result cannot tell you.
We request your home page, then follow ordinary links and your sitemap to find the pages a buyer would read: pricing, products or services, booking, contact, FAQ, about, locations and policies.
A small number of pages is read per test. Pages are ranked by how likely they are to hold buying information, not by how the site is designed.
We read what any visitor can read. We do not sign in, submit forms, create accounts, attempt payment or place bookings, and we do not try to reach private or restricted areas.
If a page cannot be retrieved, it is recorded as inaccessible rather than treated as missing information.
Language, page structure and commercial signals are used to decide whether the site is an online shop, restaurant, venue, professional services firm, local service business or software product.
Classification carries a confidence level. When confidence is low we say so, keep guidance general, and do not sell a report built on a shaky reading.
Eight categories are always scored: offer clarity, decision criteria, pricing, availability, trust, actionability, transaction and machine readability.
Their weighting adapts to the business type. Live availability matters enormously for a restaurant and very little for a consulting firm, so the same evidence produces different scores for different businesses.
Every finding is tied to a status: confirmed (we saw it), inferred (strongly implied but not stated), missing (not present on the pages read) or inaccessible (we could not reach it).
Nothing is assumed. If a price is not published, the result says the price is missing — not that the business is expensive.
Scores are produced by fixed rules, not by asking a language model for an opinion. The same pages with the same content produce the same score under the same model version.
The buyer journey in your result is a simulation built from the evidence we retrieved. It is not a transcript from ChatGPT, Gemini, Copilot or any other product, and no assistant was asked to buy from you.
Each step states what the buyer needed, what we found, whether that satisfied the requirement, where it came from and what it means for a recommendation.
Structured data, an llms.txt file, an API, published availability, pricing pages and booking or payment pathways all make facts easier for a machine to use.
None of them create facts. If eligibility, price, capacity or a completable next step is not published anywhere, machine-readable markup cannot supply it, and the score reflects that.
Websites that render their content only through JavaScript, or that block automated requests, will return less evidence than they actually contain. We report that as inaccessible rather than penalising it silently.
Sites behind aggressive bot protection may fail the test entirely. That itself is useful information: AI buyers meet the same wall.