Side by side
Your score on its own answers "am I good?", which is hard to act on because nobody knows what good is. Next to two competitors it answers "am I behind, and where?" — which has an obvious next step. Every number is a real measurement of a real page, and the result has a link you can send.
What gets compared
The overall number tells you who is ahead. The pillars tell you where, which is the part you can act on: being twenty points behind on Structure is a week of work on your product data, while being twenty behind on Reachability is usually one line in a configuration file that somebody added without knowing what it turned off.
Every report answers seven questions a customer actually asks — what this company is, what it sells, what it costs, how to reach it, where it is, whether it is in stock, what happens if they send it back. The comparison shows how many of them each site answers in a form a machine can lift out rather than paraphrase. That count is often the widest gap on the page.
Resolved from each site’s own robots.txt, crawler by crawler, for the exact path being compared. A great many sites blocked the crawlers that train on them and the ones that send them traffic in the same stroke, and the ones that did it usually have no idea. Seeing a competitor allowing what you refuse is the fastest way to find out you are one of them.
An estimate of the tokens each page occupies in a model’s context, and how much of that is navigation, header and footer rather than the page’s own content. A template that spends two thirds of every page on a mega-menu is paying that on every page an assistant ever reads, and it is the kind of thing nobody measures until they see somebody else’s number beside theirs.
Everything here is a measurement of a page that was fetched, at the moment it was fetched, with no browser and no JavaScript — the same budget a crawling agent gives you. Nothing is modelled, estimated from a sample, or averaged across a site. The result gets a permanent address you can send to somebody, and the numbers on it never change afterwards, because they are the record of one measurement rather than a live dashboard.
Comparing somebody else’s public page is an ordinary thing to do — it is what a browser does — but scanning is not publishing: nothing you compare enters the public AX Index unless that domain’s owner proved control and asked to be listed. A site that would rather not be fetched by us at all can say so in two lines of robots.txt, and that is honoured here as everywhere else.