Each line scores +2 for a precise number, +2 for a named method (n=, sample, margin of error), +2 for attribution (said, according to), -1 for a hedge (most, many, significant), +1 for fitting a chunk (45 words or fewer). 5+ is citable. A research hub should clear a density of 4 per 1,000 words.
Per-claim verdicts appear here.
What citable stat density measures
A generative engine does not lift your study. It lifts a stat unit: the smallest self-contained block carrying a precise number, a named methodology, and an attributed quote. Citable stat density counts how many of those extractable units your document holds per thousand words. A general blog post scores under 1. A well-built research hub clears 4. Below 2 you have written an essay with numbers in it, which a model can read but will not confidently repeat.
How to use it
- Paste your draft, one claim or sentence per line.
- Read the per-line verdict: CITABLE, WEAK (add a method or attribution), or REWRITE.
- Fix the WEAK and REWRITE lines first, usually a missing methodology or a hedge word that should be a number.
- Watch the density figure climb as you tighten claims. Aim to clear 4 per 1,000 words before the study ships.
What is a good citable stat density?
Clear 4 extractable units per 1,000 words for a research hub. A general blog post scores under 1, and below 2 the document reads as narrative with numbers rather than a set of liftable facts. The number matters less than the per-line verdicts: every WEAK or REWRITE line is a finding a model cannot confidently quote.
How is each line scored?
Plus 2 for a precise number, plus 2 for a named method (n=, sample, surveyed, margin of error), plus 2 for attribution (said, according to, per, told), minus 1 for a qualitative hedge (most, many, significant, leading), and plus 1 if the line is 45 words or fewer so it fits a retrieval chunk. Five or more is citable; three to four is weak; below three needs a rewrite.
Why do hedge words lower the score?
Because a model cannot repeat most teams or a significant share as a fact. Retrieval weights precise, verifiable claims and discounts qualitative filler, so a hedge is a missed stat unit. Replace the hedge with the number it is standing in for, and add the methodology that makes the number trustworthy.
Does my draft leave the browser?
No. Every line is scored entirely in your browser with simple pattern matching. Nothing you paste is uploaded, stored or sent to any server.