When we get it wrong
Every score on this site comes from a deterministic pattern matcher reading one HTML document. That is fast, free and repeatable, and it is also why the scanner is sometimes confidently, specifically wrong about a page somebody spent months on. This is what happens when you tell us so.
How to argue with a result
Email us. That is the whole mechanism, and it is deliberate. There is no account to make, no form that files a ticket, and no queue holding your name and address — this site scans other people's pages and stores nothing, and a database of complaints would have been the one place that stopped being true.
Send the URL and, if you can, the specific finding you think is wrong. The signals reference names every detector and says exactly what it matches on, so “invented-stats fired on a figure we actually measured” is worth ten emails saying the score felt harsh. Naming the detector is the difference between a complaint and a bug report.
What happens next
- A person opens the page and reads the markup. Not a model, not a rescan. The question is always the same: does the detector match what it claims to match, on this document?
- One of three things is true. The detector is right and the page really does have the thing. The detector is right but the thing is fine in context — which is a scoring problem, not a matching one. Or the detector is wrong, and that is the good outcome, because it can be fixed for everybody.
- Something changes, or it does not. Most of these end privately: a threshold moves, a pattern gets narrower, or we conclude the result was correct and say so in a reply. A detector change ships as a commit and shows up in the signals reference.
- A few get written up. Only when the case teaches something, and only with the permission of the person who raised it. Those are below.
What we can actually verify
Very little, and pages here say so rather than implying otherwise. We cannot confirm that you own a domain, that you built the site, or that the person emailing is who they say they are. So a written-up case says “raised by a person identifying as the developer” unless somebody verified it by hand and wrote down how — in which case the page says what the method was.
The evidence a case does rest on is the part that is checkable: the URL, the markup, the detector, and what it matches. That is public, and you can re-run it yourself on the scanner.
Written-up cases
None yet. That is the honest state of it — and an empty grid of placeholder cards would be precisely the thing this site exists to point at, so there isn't one.
When there is a case worth reading, it will be here: the claim, the markup, the detector, what we found, and what changed in the scanner as a result.