ReadabilityProbe
Paste a page’s HTML. Sixteen weighted structural checks report how well it survives machine extraction, and what each failure costs. Everything runs in your browser; nothing is uploaded.
…
Method
What it checks,
and why
Sixteen checks in five groups. Each carries a weight, because a missing H1 and a missing canonical are not the same size of problem.
- Identity · weight 7
- Title, meta description, declared language, canonical URL. The labels a page puts on itself before anything reads a word of it.
- Outline · weight 7
- A single H1, contiguous heading levels, and words-per-heading. Passage extraction leans on the heading hierarchy; a broken outline produces orphaned chunks.
- Structured data · weight 3
- JSON-LD present and parsing. A block that throws is worse than no block at all, because it is discarded without complaint.
- Semantics · weight 3
- Landmark elements, and the ratio of generic divs to meaningful ones. Boilerplate removal has to guess which part is the content; landmarks are how you stop it guessing.
- Content shape · weight 8
- Extractable word count, text-to-markup ratio, described images, descriptive link text, question-shaped headings, and the length of the opening paragraph, the passage most likely to be quoted.
This is a structural read, not a quality judgement. A beautifully structured page of nonsense scores well; a brilliant essay in one unbroken block scores badly. It also sees only the HTML you give it, so a page that assembles its content with JavaScript will look emptier here than it does in a browser, which, for the crawlers that do not execute scripts, is accurate rather than a bug.
The weights are a first guess. They get revised when EXP-003 starts producing evidence about which structural traits actually correlate with being cited.
Where your input goes
Nowhere
The probe parses your HTML into a detached document inside your browser and reads it there. No request is made and nothing is stored. The HTML you paste never leaves the tab. The only other calls this page makes are the webfont request, the site analytics, and the optional button that fetches an article from this same site.