Fraction of factual assertions in publicly accessible web text that are false
- Measures
- The proportion of factual assertions (checkable statements of fact, excluding opinion, fiction, and prediction) that a careful fact-checker would judge false at the time of assessment. Each assertion counts once, not weighted by readership.
- Population
- Factual assertions in publicly accessible web text, as represented by a general-purpose web crawl such as Common Crawl
- Conditions
- Unweighted random sample of assertions from current crawl; outdated-but-once-true statements counted as false only if presented as current
- Unit
- a share, shown as a percentage
Between 1 and 30% (90% interval), most likely about 5%.
Stated as an interval, the precision the evidence supports. right-skewed; strongly sensitive to how assertions are counted and what counts as false
Assessment
The graph's estimate, all things considered: the figures and arguments it rests on, and how they were combined.
No study has measured this fraction directly. The definition fixes an unweighted random sample of checkable factual statements from a general web crawl, judged by a careful fact-checker. Under that definition the value is most likely a few percent, with a central value near 5%. Most crawled text is commercial, navigational, reference, or routine reporting, and these kinds of text have low error rates. Sites that rate low on credibility account for a small share of content. Forums, search-optimised filler, and partisan pages raise the rate. The 90% interval of roughly 1% to 30% is wide because there is no direct measurement, and because the value depends heavily on how assertions are segmented and on whether outdated or imprecise statements count as false. The figure is a reasoned judgment, not a measurement.
The work: which figures, how they were combined, and the arithmetic
This reassessment confirms the previous estimate. Its budget allowed no new searches, so no new figures were recorded. The estimate rests on three background considerations. (1) Domain-level credibility ratings and studies of misinformation exposure put low-credibility sources at a single-digit percent of web content and traffic. (2) Accuracy comparisons of curated reference text find error rates of a few percent of statements or less. (3) User-generated and low-quality commercial text is less accurate. Combined, these give a central value near 5% and a long right tail. What would sharpen it: a fact-checking audit of a random sample of assertions drawn from a public crawl such as Common Crawl, with the rules for segmenting assertions and judging falsity fixed in advance.
Decomposition
This claim is atomic: it bottoms out in a bedrock fact, an open empirical question, or a value premise, and does not decompose further.
Cite this claim: a formal citation with its evidence attached
Contribution record · 1 exchange
Challenges and proposals from contributors, with the review each received. Substantive exchanges stay on the claim's public record.
Parent question for estimates of how much of what is said online is false. It is widely asserted that much online content is false, but there is no agreed population of "claims on the internet," and most available figures come from fact-checker samples that are selected for suspicion. The question may prove ill-posed as stated; if so, the useful work is decomposing it into well-defined quantities (random samples of web pages, viral social media posts, fact-checked claims) and saying how each bears on it.
The proposal asks what fraction of factual assertions in publicly accessible web text are false. This is a neutral, well-formed question that anyone reasoning about online misinformation would need to ask (§7a). It concerns no private person. It is a quantity-style question (how much), so the Matcher can enter it as the quantity it asks about, or as a question if its definition proves incomplete. The contributor flags that the population ("claims on the internet") is undefined and that fact-checker samples are selected for suspicion. That is a reason to map the question carefully, not to exclude it, and the Steward can decompose it into better-defined quantities (random web samples, viral posts, fact-checked claims) as suggested. Novelty and any wording refinement are for the Matcher and Steward. The supporting note is sincere and on-topic.
Contribute
Every judgment on this page is open to challenge. A contribution is evaluated on its merits by the reviewer; if it succeeds the page changes, and if it does not, the reasons are stated. Either way the exchange becomes part of the claim’s public record.
Created by user · Nov 2, 2026. Every judgment on this page is accompanied by a reasoning trace.