Content analyzer
What is fact density and why does LOW_FACT_DENSITY reduce AI citability?
Seeklyn's LOW_FACT_DENSITY flag fires when fewer than 1 fact token appears per 100 words — a fact token being any digit, decimal, percentage, 4-digit year, or dollar amount matched by the pattern \b(\d+([.,]\d+)?%?|\d{4}|percent|\$\d+)\b. Fact density is the ratio of concrete data points to total words in a passage.
Seeklyn flags this in every content analysis — run your page free at /dashboard/content.
What is fact density?
Fact density is the count of concrete data points divided by total word count, expressed as facts per 100 words. Seeklyn fires LOW_FACT_DENSITY when that ratio falls below 1 — meaning the passage has fewer than 1 fact token for every 100 words it contains. The threshold is intentionally minimal: a single number, year, or percentage per 100-word block is all that is required to clear the flag.
A fact token is any of the following: a digit or decimal (42, 3.5), a percentage (87%), a 4-digit year (2024), the word “percent” spelled out, or a dollar amount ($15). Vague quantifiers like “many,” “most,” or “significantly” do not match the regex and do not count. The before/after example illustrates the difference: “Many companies have seen improvements in organic search performance after optimizing for structured data” has zero fact tokens. “Structured data adoption lifted AI Overview inclusion by ~15% for FAQ-schema pages, per Google Search Central data from 2024” has three (15%, 2024, and an implicit count in the percent figure) and clears the flag easily.
Why does low fact density reduce AI citability?
AI answer engines extract specific, quotable answers to relay to users. A passage that offers only vague generalities gives the engine nothing concrete to attribute or quote — it contains no number to cite, no year to verify, no figure to anchor a claim. LOW_FACT_DENSITY is a moderate-evidence flag in Seeklyn's rubric, meaning the correlation between factual specificity and citation is well-observed across content audits, though not yet peer-reviewed at the level of structural flags like ANAPHORA_OPEN.
Beyond the technical flag, the underlying dynamic is that AI engines model authoritativeness. A passage stuffed with hedges and no data reads as speculative; a passage with even 1 concrete figure reads as informed. Passages that clear LOW_FACT_DENSITY are more likely to be surfaced by Google AI Overviews, Perplexity, and ChatGPT as supporting evidence for a query answer.
How to increase fact density
Adding numbers, dates, percentages, or dollar amounts is the direct fix. One concrete datum per 100 words is the floor Seeklyn enforces — aim for 1–3 substantive data points per passage rather than front-loading all figures into one sentence. Replace phrases like “some improvement” with “a 12% improvement (HubSpot, 2024)” and “recently” with the actual year.
Appropriate sources make the figures more credible and also satisfy the companion NO_ATTRIBUTION flag. A fact token alone clears LOW_FACT_DENSITY; a fact token with an attribution phrase (“according to,” “per,” or a year in parentheses) clears both flags at once and raises the citability score on two dimensions simultaneously.
Frequently asked questions
What counts as a fact token for Seeklyn?
Seeklyn's fact-token regex matches: any digit (including decimals), any percentage, 4-digit years (e.g., 2024), the word “percent” spelled out, and dollar amounts. A year like “2023” counts. A vague phrase like “many” or “most” does not.
Does a year alone satisfy fact density?
Yes — a 4-digit year (matching \d{4}) fires the fact-token regex. A passage that contains at least 1 year per 100 words clears the LOW_FACT_DENSITY threshold. Including the year is also the minimum attribution signal that clears NO_ATTRIBUTION.
Is more always better for fact density?
Not automatically — extremely high fact density can trigger KEYWORD_STUFFING if the same term repeats unnaturally alongside the figures. Aim for 1–3 substantive data points per 100-word passage: enough to be concrete, not so many that the prose reads as a data dump.
Can a passage with a high citability score still fire LOW_FACT_DENSITY?
Yes — clearing LOW_FACT_DENSITY alone does not guarantee a high score; all 8 flags are independently checked. A passage might answer first and name entities (clearing ANAPHORA_OPEN and ENTITY_SPARSE) while still lacking any numeric data, which fires LOW_FACT_DENSITY regardless of the other signals.