Paste a passage below and this tool scores how quotable it is by AI search engines. It checks the same signals that decide whether ChatGPT, Perplexity, or a Google AI Overview lifts your text as a cited source: length, information density, self-containment, and whether it answers the question directly. Everything runs in your browser. Nothing is sent anywhere.
This is a heuristic guide, not a guarantee. It scores structure and specificity, the parts you control; it cannot judge whether your facts are correct or whether your page ranks well enough to be retrieved in the first place. Use it to tighten a passage, then do the rest of the citation work.
What each signal means
The score weighs four things, drawn from the patterns in our AI search citation benchmark:
- Length. A single liftable passage tends to sit between 40 and 160 words. Shorter than that rarely answers a question on its own; much longer and the engine has to hunt for the answer inside it.
- Information density. Named entities and concrete numbers per sentence. "Frontier models encode 95 to 98 percent of the facts tested but cannot always retrieve them" gives an engine something to ground on; "the recent update had an impact" does not.
- Self-containment. The passage should read on its own without the sentence before it. Opening with a bare "This" or "It" points at context the engine will not have when it lifts the chunk.
- Direct-answer shape. Lead with the answer. A definitional opening ("X is...", "X works by...") gets lifted more cleanly than a passage that hedges its way in.
How is the score out of 100 calculated?
The score is a weighted average of the four signals above: length contributes 30%, information density 25%, self-containment 25%, and direct-answer shape 20%. It runs entirely in your browser, so nothing you paste leaves your machine. Above 85 it reads "Highly citable", 65 to 84 "Citable with tightening", 45 to 64 "Needs work", and below 45 "Unlikely to be cited". None of the four is about writing quality in the ordinary sense. Every one of them is about whether a passage survives being cut out of your page and dropped into an answer with no context attached.
Why 40 to 160 words?
A passage between 40 and 160 words scores full marks for length because that is the band where a chunk can answer a question completely and still be lifted whole. The scorer awards 70 for 25 to 39 words, 65 for 161 to 300, 40 above 300, and 35 below 25. Under about 25 words a passage rarely finishes an answer, so an engine has to reach for the paragraph after it and the citation goes elsewhere. Past roughly 160 the answer is in there, but it is surrounded by material the engine has to discard, and discarding is where a competing source that needs no editing wins.
What counts as information density here?
Density is counted as the number of concrete figures plus capitalized named entities in a passage, divided by the number of sentences. At 1.2 or more per sentence the signal scores 100; at 0.6 to 1.19 it scores 75; below 0.6 it scores 40. So a four-sentence passage needs roughly five named things or real numbers to clear the top band. This is deliberately mechanical, because the property it approximates is mechanical: a model grounding an answer prefers text carrying something checkable. "Frontier models encode 95 to 98 percent of the facts tested but cannot always retrieve them" gives it that. "The recent update had an impact" gives it nothing to hold on to.
What makes a passage self-contained?
A passage is self-contained when its first word names its own subject rather than pointing at the sentence before it. The scorer checks the opening word against ten referring words: this, that, these, those, it, they, them, here, there and such. Finding one drops the signal from 100 to 45. That single check catches most of the failures, because a paragraph beginning "This means the threshold moves" is complete on your page and meaningless the moment it is lifted. The same logic drives the answer-shape signal: an opening that states the answer plainly scores 100, a hedge such as "it depends" or "there are many" scores 45, and anything in between scores 70.
How is this different from a readability score?
A readability score is a measure of how easily a human reads a sentence. Flesch-Kincaid, the best known, works from average sentence length and syllables per word, and penalizes the long technical vocabulary this scorer rewards at its 1.2 density threshold. The two tools measure close to opposite properties. Flesch-Kincaid asks whether the reader in front of you can follow the paragraph; this asks whether the paragraph survives being cut out of your page entirely. A dense passage naming Google, Perplexity and a real percentage can score badly on Flesch-Kincaid and 100 here, and both results are correct. Readability is about the reader you already have. Citability is about whether an answer engine sends you a new one.
Where do these thresholds come from?
Every threshold on this page is ConceptSEO's rather than Google's, and it is worth being plain about that. Google's Search Central documentation on AI features and your website states that the same foundational SEO practices apply to AI features as to Search overall, and that no new machine-readable file, AI text file or special schema.org markup is needed to appear in them. It publishes no preferred passage length and no density target. The one place it does get specific is the controls that suppress a quote: nosnippet, data-nosnippet, max-snippet and noindex all limit what can be shown from your page. The four bands above come from running these same checks across the domains ConceptSEO manages every week and watching which passages get quoted.
Why this matters
Answer engines do not retrieve pages, they retrieve passages. A long article can go uncited while one well-formed paragraph on a thinner page gets the mention. Tightening the passage is the highest-leverage thing you can do on-page, and it is the same work behind our GEO service. Score a passage, fix what the tool flags, and you have made it easier for an engine to quote you by name.
If you want a read on where your whole site stands in AI answers rather than one passage, a free audit is the place to start.