How the scores work
Every reading runs 0-100 and higher is always better. The bands below are the published ladders, read straight out of the same constants the scorers use.
One scale, three vocabularies
Three tools reading one site would otherwise produce three numbers a reader has to reconcile. They share the scale so the numbers can be compared, and keep their own band names because what a number means is not the same in each.
A consolidated verdict is the worst of the readings that ran, not an average. An average hides the one thing that is actually wrong.
Can an answer engine read the page?
| Band | Score | What it means |
|---|---|---|
| Legible | 81-100 | A machine reading this page gets the whole of it. |
| Patchy | 61-80 | A machine gets most of the page and works around the rest. |
| Obscured | 41-60 | A machine gets fragments, and infers the rest of the page for itself. |
| Unreadable | 0-40 | A machine gets nothing it can use from this page. |
Any blocker pins the score into the bottom band on its own. A page that says noindex, or serves no text without JavaScript, or answers a crawler with 403, cannot be quoted whatever else is true of it.
What is the site exposing?
| Band | Score | What it means |
|---|---|---|
| Sealed | 81-100 | Nothing obvious is leaking from the browser. |
| Exposed | 61-80 | A secret or a misconfiguration is visible in what the site ships. |
| Cracked | 41-60 | Readable data or a live key. Fixable today, dangerous until it is. |
| Wide Open | 0-40 | Anyone with the URL can read this app's data right now. |
How much came out of a template?
| Band | Score | What it means |
|---|---|---|
| Hand-Crafted | 81-100 | Almost nothing here comes out of a box. |
| Lightly Templated | 61-80 | Mostly deliberate, leaning on a few defaults the crowd also uses. |
| Heavily Templated | 41-60 | Untouched defaults across copy, layout and type. |
| Pure Slop | 0-40 | Stock everything. The template is still showing. |
Calibrated against pages that are demonstrably quoted
The citation weights are set by a corpus of pages answer engines quote every day: Wikipedia, MDN, and the Python, React, nginx and Postgres references, alongside twenty popular pages across news, retail, health, government and SaaS. The rule fixing the numbers is that such a page must come out in the top band, because nothing is in fact standing in its way.
That rule is what makes the score usable. Before it, the nginx reference scored 68 about a plain served HTML page that is quoted constantly. A scoring system that fails the things everyone cites is measuring its own opinions.
Why an absence costs nothing
A signal a page simply does not publish is listed and weighs zero. Pricing absences built a score out of choices the most-quoted pages on the web make and are quoted anyway.
A broken attempt is different and keeps its weight. Somebody meant that canonical to work, and it does not.
Read a site with Readout