Weights: 0.30 position, 0.30 density, 0.20 headings, 0.20 neutrality. Density is capped at 1.0 once the page averages 3.0 facts per 100 words. No engine exposes its weights, so treat this as a prioritisation heuristic: fix the lowest term, then re-score.
Your weakest lever, and the fix, appears here.
Why score extractability, not rank
A comparison page is not retrieved as a page, it is judged one passage at a time. This score rolls the four levers that decide whether those passages get quoted, answer position, fact density, question-style headings and neutrality, into a single 0 to 100 number so you can prioritise a backlog. The weights are a working heuristic, not a published ranking function. What matters is that all four get measured, because teams reliably optimise the one they find least uncomfortable and ignore the other three.
How to use it
- Set answer position: is a verdict naming both products inside the first 100 words?
- Enter fact density, the count of verifiable facts (numbers, prices, dates, named standards) per 100 words.
- Enter the share of your H2s phrased as the buyer's actual question.
- Set neutrality: a named competitor advantage plus two or more outbound citations earns full marks.
- Read the weakest lever, fix that one, and re-score.
Is this an official Google or OpenAI ranking score?
No. No engine publishes its retrieval weights. This is a practitioner heuristic that bundles four levers with strong experimental and observational support into one comparable number, so you can rank pages by how much retrieval work they are leaving on the table.
Why is fact density capped at 3.0 per 100 words?
Because returns flatten. A passage that runs one verifiable fact per 100 words has crossed from decoration into evidence; comparison pages that run 2.0 to 4.0 already sit in the range models reward, so the term maxes out at 3.0 to avoid over-rewarding stat-stuffing.
Does my data leave the browser?
No. The calculation runs entirely in your browser. Nothing you enter is sent to a server or stored.
What score should I aim for?
Treat 70 and above as retrieval-ready and below 40 as at risk, but the useful signal is the weakest term, not the total. A page at 55 with zero fact density will move further by adding evidence than by polishing anything else.