What is fact density?

Fact density is the proportion of a text made up of specific, checkable factual claims — names, dates, numbers, citations — versus opinions, generalities, and vague statements. High fact density means more to verify and more places an error can hide.

How does the tool classify sentences without an internet connection?

It uses linguistic heuristics — the presence of numbers, dates, proper nouns, and citation cues marks verifiable claims; opinion verbs and hedging mark interpretations; confident specific claims with no source marks hallucination risk. It does not check facts against the web, so it flags what to verify, not whether a claim is true.

Can it tell me if a claim is actually correct?

No. It cannot verify truth — that requires sources. It identifies which sentences make checkable factual assertions and which carry elevated hallucination risk, so you know where to point your own verification effort.

Why are confident-but-unsourced specifics flagged as risky?

Hallucinations most often appear as precise, confident details — a specific statistic, a named study, an exact date — presented without a source. The combination of high specificity and zero attribution is the strongest text-only signal that a claim warrants checking.

Is my text sent anywhere?

No. The analysis runs entirely in your browser using local heuristics. Nothing you paste is transmitted to us or any third party.

What is the AI Output Fact Density Analyzer?

Paste AI-generated text and get a fact-density analysis — classifying sentences as verifiable factual claims, opinions or interpretations, and hallucination-risk statements, based on linguistic confidence markers and claim structure. It runs free in your browser on Gera Tools, with nothing uploaded.

AI Output Fact Density Analyzer

Name: AI Output Fact Density Analyzer
Creator: Gera Tools
License: https://creativecommons.org/licenses/by/4.0/

AI output fact density analyzer

When you ask a model for an answer, the dangerous part is not the obvious opinion — it is the confident, specific-sounding claim with no source behind it. This analyzer measures the fact density of AI-generated text: it classifies each sentence as a verifiable factual claim, an opinion or interpretation, or a hallucination-risk statement, and reports the overall mix. It cannot tell you whether a fact is true — only a source can — but it tells you exactly where to aim your verification.

How it works

The tool splits the text into sentences and scores each with local linguistic heuristics. Numbers, dates, proper nouns, and citation cues mark a sentence as a verifiable factual claim. Opinion verbs (“I think”, “arguably”) and value words mark an opinion or interpretation. A sentence that is highly specific and confident yet carries no source or hedge is flagged as hallucination risk — the classic shape of a fabricated detail. It then reports counts and a fact-density score. Everything runs in your browser.

The three sentence categories explained

Verifiable factual claim — A sentence that asserts something checkable: a named entity, a date, a number, a statistic, a stated outcome. These sentences are the ones you can verify against a source. A high count here is not bad; it means the response has substance. It also means more fact-checking work.

Opinion or interpretation — A sentence that expresses a judgement, a recommendation, or an analysis. “This is likely the best approach” or “the key implication is…” These are where the model is reasoning, not retrieving. They can be valuable, but they are not verifiable in the same way and should be read as the model’s perspective rather than established fact.

Hallucination risk — The dangerous category. A sentence that is highly specific and confident — it reads like a factual claim — but carries no source, citation, or hedge. “The company was founded in 1987 and acquired in 2019 for $240 million” is a hallucination-risk sentence if there is nothing in the prompt to ground those figures. Precision without attribution is the canonical shape of a fabricated detail.

A worked example of why density matters

Consider two AI-generated paragraphs on the same topic. Paragraph A consists mostly of interpretations: “AI adoption is growing rapidly and is expected to transform many industries.” Low fact density, low hallucination risk, but also low verifiability — there is nothing to check. Paragraph B includes: “According to a 2023 survey, 67% of enterprises reported using AI in at least one business function, up from 45% in 2020.” Higher fact density — the specific percentages and years are checkable — but also higher hallucination risk because those exact figures could be invented. Running the analyzer prompts you to verify those numbers, which you would otherwise read past.

Tips for effective use

Run it before publishing AI-assisted content to identify which sentences need a source before they go live.
Pair with a citation prompt. Ask the model to add citations, then re-run — attributed claims drop out of the risk bucket.
Higher density in niche topics warrants more scrutiny. Models are more likely to confabulate precise details in specialised domains where training data is sparser.
Heuristics, not magic. Rhetorical numbers (“a thousand different options”) and quoted opinions can trigger false classifications; treat the output as a guide for where to focus, not a definitive verdict.