Measuring
The extractability score
The score is 0–100 and it is coarse on purpose. It ranks pages for attention; it is not a precise measurement of anything.
What makes up the score
| Component | Points | What it measures |
|---|---|---|
| Not thin | 20 | The page has at least 300 words. Below that an engine has too little to work with. |
| Has an answer block | 25 | The first paragraph is between 10 and 120 words — long enough to answer something, short enough to lift whole. |
| Has a modified date | 10 | The page exposes when it was last updated. |
| External citations | 7 each, up to 20 | Distinct external hosts linked. Self-links corroborate nothing, so they do not count. |
| Statistics | 5 each, up to 15 | Percentages, currency amounts, scaled counts and multipliers — the specific figures that get quoted verbatim. |
| Question headings | up to 10 | Proportion of your h2/h3 headings phrased as questions, which is how people actually prompt an engine. |
The weighting is deliberate. The answer block carries the most of any single component because it is the one an engine cannot work around, and citations plus statistics carry 35 between them because they have the strongest published support behind them.
What the score does not mean
- It is not a prediction. A page scoring 90 is more extractable than one scoring 40. Neither number says anything about whether an answer engine cited either.
- A score change is not a visibility improvement. It is a content-structure measurement that changed. Crawlune will not describe it as anything else.
- It is not precise. It is never shown to a decimal place, because the underlying signals do not support that precision.
Where it under-reports
Worth knowing, because they explain scores that look unfair:
- Page-builder content. Crawlune reads stored post content. A site built in Elementor, Divi or WPBakery keeps much of its text in ways that reading is not, so such a site can score lower than it deserves.
- Non-English content. Some signals — question headings in particular — are detected with English-language patterns, so a non-English site will under-report on them.
These are limits of the measurement, not verdicts on the page. Crawlune states them rather than quietly scoring you down.