Kuroma
Citation-Worthiness
What this check reads on your page
Claims an engine could repeat as fact are what the visible text is read for here. The check counts statistics: percentages, money amounts in any of the common currency forms, and thousands-separated figures. It counts quotations, both blockquote elements and passages set inside matched quotation marks of any family, 「」 included. It looks for English phrases that signal original data ('our research', 'we found') and that flag an insight ('key takeaway', 'notable'), and it measures hedging, deducting a little when 'might', 'may' or 'possibly' appear in more than eight of every hundred sentences.
Our verdict on the evidence
Evidence-backed Verifiable evidence units (statistics, quotes, sourced claims) causally improve citation in sandboxes, and confident wording is preferred over hedged wording. Effects shrink in competitive settings, so claims stay capped.
GEO: Generative Engine Optimization (arXiv 2311.09735) · Cross-engine citation-driver study, deep vs shallow content and freshness gatekeeping (arXiv 2605.25517)
A figure counts here whether or not it is true: the check reads the shape of a claim, not its accuracy, so it can tell you that a page asserts nothing checkable but not that what it asserts is right. What it stands in for is what an answer engine does with a sentence. A number, a named source or a direct quotation can be lifted into an answer as evidence, while a sentence that says something 'may' be the case gives the engine nothing to attribute and is the first thing it paraphrases away.
What it is worth
25% of AI Readiness, which is 7.5 points of the overall 100.
On a page written in Japanese, Korean or Traditional Chinese, only the parts of this check that work in any language are scored, over a budget of 50 rather than 100, and the result is rescaled to 100 so it stays comparable.
How audited pages score on it
Based on 8,875 audited pages.
The median page scores 8, and the middle half of pages score between 0 and 22.
- Good (75 to 100)32 (0%)
- Moderate (50 to 74)522 (6%)
- Needs work (25 to 49)1,582 (18%)
- Critical (0 to 24)6,739 (76%)
94% of audited pages score below 50 on it.
Where it fails most and least
Industries that fail it most
- Gaming: 100% of 213 audited pages fail it
- Marketplace: 100% of 46 audited pages fail it
- Healthcare Services: 100% of 45 audited pages fail it
- Government / GovTech: 100% of 36 audited pages fail it
- Pharma / Biotech: 99% of 137 audited pages fail it
Industries that fail it least
- MarTech / Marketing Platform: 83% of 110 audited pages fail it
- Construction Tech: 86% of 42 audited pages fail it
- HR Tech / Recruitment: 86% of 71 audited pages fail it
- Supply Chain / Logistics: 87% of 63 audited pages fail it
- Cybersecurity: 87% of 69 audited pages fail it
How to fix it
- State the number: a percentage, a price in a named currency, a count with its thousands separator.
- Publish something you measured yourself and say so in the phrases the check reads: 'our survey found', 'we found', 'original data'.
- Quote a named person or source in a blockquote, or set the quotation inside quotation marks.
- Replace 'might' and 'may' with the specific claim you can stand behind; hedged sentences count against the page.
- Flag the finding in words an engine can spot, at the start of the sentence that carries the number: 'key takeaway', 'notable', 'significantly'.
What has changed in this factor
No version names this factor in its per-factor record.