Reliability of Wikipedia
Reliability of Wikipedia is the question the entire experiment hangs on: can an encyclopedia that anyone can edit be trusted? The honest answer, supported by two decades of studies and scandals alike, is a conditional one — Wikipedia's reliability is not a property of the site but a property of the individual article, and it varies with traffic, topic, and the maturity of its sourcing in ways readers can learn to spot. This article surveys what the research actually found, where the model reliably fails, and how to read it safely. For the broader catalog of scandals and structural complaints, see Criticism of Wikipedia.
The study that set the terms
The founding measurement is Nature's December 2005 blind comparison: reviewers drawn from the scientific community compared 42 pairs of Wikipedia and Encyclopaedia Britannica science articles without knowing which was which, and counted errors. The result — 162 errors in Wikipedia's set versus 123 in Britannica's, roughly four per article against three — astonished an establishment that had assumed the volunteer project would be laughably worse. Britannica disputed the methodology (calling some counted "errors" matters of emphasis and omission rather than factual mistakes); Nature stood by the study. Whatever the margin, the finding fixed the terms of the debate ever since: Wikipedia's science coverage is not perfect, but it belongs in the same conversation as the gold standard rather than a separate, inferior category.[1]
What later research found
Subsequent studies refined rather than overturned that picture:
- A 2014 systematic review of prior reliability studies (Journal of the Association for Information Science and Technology) found Wikipedia's accuracy generally good on well-developed articles but uneven across the corpus — the recurring finding being variance, not uniform failure.
- Health-information studies are split: several reviews found drug and medical articles broadly accurate but incomplete — reliably right about what they cover, thin on dosing interactions, rare side effects, and specialist nuance. Medical professionals' consistent advice — use it to orient, not to self-prescribe — mirrors the academic consensus for all topics.
- Legal and historical accuracy fares worse under scrutiny: a frequently cited 2010 study of U.S. Supreme Court coverage found persistent inaccuracies lingering in a majority of the cases examined, and humanities topics generally attract fewer expert eyes than STEM.
- Comparative studies consistently find accuracy correlates with article development level — featured and good articles (the internal peer-review badges) are close to professional reference quality; stubs and low-traffic pages are not.
The variable that predicts accuracy: traffic
The strongest single predictor of an article's reliability is how many eyes it gets. High-traffic articles — heads of state, current events, popular science — are watched by hundreds or thousands of editors; vandalism there is typically reverted within minutes, and outright errors rarely survive a news cycle. Low-traffic articles live under a different physics entirely: an obscure village, a forgotten general, a technical mineral's properties — if the initial author was wrong, correction waits for a knowledgeable stranger to happen by. Every famous hoax in Wikipedia's history — Seigenthaler, the Brazilian aardvark, the Bicholim conflict, Jar'Edo Wens, the MacMasters toaster — survived precisely because it lived in the long tail.[2]
The reader's translation is simple: popularity is a proxy for scrutiny. An article with a hundred watchers is a reviewed document; an article with three is a manuscript.
Vandalism and repair speed
Crude vandalism — blanking, profanity, joke edits — is the least serious threat, because it is the most visible: recent-change patrols, edit filters, and anti-vandalism bots (ClueBot NG, the most famous, reverts within seconds) suppress it fast on watched pages. The dangerous failures are the quiet ones: a subtly altered date, a plausible-sounding fabrication, a PR-shaped paragraph — edits that look like contributions and are reviewed only by whoever happens to compare them against sources. Studies measuring vandalism lifespan consistently find a bimodal distribution: obvious damage dies in minutes; sophisticated damage can persist for years.
The citation machinery — and its failure mode
Wikipedia's core content policy is verifiability, not truth: an article may state only what published reliable sources have stated — which means reliability is, in practice, delegated outward to journalism and academia. The strengths are real: a well-cited article is an annotated bibliography with prose attached, and the citations let readers audit the claims. The failure mode is circular reporting — what the community calls citogenesis: a journalist copies a claim from Wikipedia, Wikipedia then cites the journalist's article as the source, and an error manufactures its own evidence. The Brazilian-aardvark hoax is the canonical case: the fake nickname propagated into newspapers, and the article eventually cited one of those newspapers as proof.[3]
A second structural weakness is the reliable-sources rule itself: on politicized topics, the outcome of a dispute is determined less by what is true than by which sources count as reliable — a judgment made by the same volunteer community whose composition shapes every other bias on the site.[4]
By topic area
- Science and mathematics: generally strong — the Nature baseline plus a self-selected editor population dense with technical professionals. Errors cluster in fast-moving subfields and in pages with few watchers.
- Current events: surprisingly good under pressure — breaking-news articles are written and corrected in real time by crowds, though early versions of developing stories carry the fog of the moment.
- History and humanities: more uneven — interpretation-dependent fields expose the consensus mechanics; minority scholarly views can be squeezed out by the weight of mainstream sources.
- Medicine: accurate in content but incomplete in coverage; Wikipedia's own medical disclaimer warns against relying on it for health decisions, and the profession's guidance is the same.
- Biographies of living persons: formally protected since the Seigenthaler affair — stricter sourcing rules, faster review — yet still the category most vulnerable to defamation, PR editing, and the reputational stakes that motivate paid manipulation.[5]
- Obscure and long-tail topics: the weakest surface — the hoax catalog lives here.
Wikipedia as infrastructure
Reliability matters beyond the site itself because Wikipedia has become infrastructure: it feeds Google's knowledge panels, trains voice assistants' factual answers, seeds the training corpora of large language models, and anchors the "authoritative source" tier of search-engine ranking. An error on Wikipedia no longer stays on Wikipedia — it replicates through the information ecosystem that treats the encyclopedia as ground truth, then occasionally flows back as a citation, completing the citogenesis loop at industrial scale. The site's defenders note the same point cuts the other way: Wikipedia is also the most scrutinized text in history, the only encyclopedia whose errors are corrected in public.
How to read it
The consensus advice — endorsed by critics and defenders alike:
- Use it to orient, not to adjudicate. Wikipedia is the world's best first stop and a poor last word.
- Read the footnotes, not the prose. The article is a summary of its sources; clicking through is the actual fact-check.
- Check the article's development level. Featured/good badges, talk-page activity, and recent-edit counts signal scrutiny.
- Distrust the long tail. Any claim on a quiet page — especially biographical or promotional — is a rumor awaiting verification.
- For anything that matters — health, law, money, citation — go to the sources Wikipedia cites, not to Wikipedia.
Verdict
Measured against the impossible standard of a flawless reference work, Wikipedia fails. Measured against every real alternative — Britannica's scope, a search engine's noise, a textbook's lag — it performs well beyond what its edit-anyone model predicts, and measurably better than it did two decades ago. Its reliability is a gradient, not a switch: high on watched, mature, technical articles; low on obscure, contested, and promotional ones. The skill it demands of readers — checking provenance — is arguably the skill the whole modern information environment demands anyway.
See also
- Criticism of Wikipedia — the full catalog of structural complaints
- List of Wikipedia controversies — the incidents behind the rules
- Seigenthaler incident · Bicholim conflict · Jar'Edo Wens · Alan MacMasters toaster hoax
- Systemic bias · Wikipedia gender gap · Ideological bias on Wikipedia
- Edit war · Arbitration Committee · Wikipedia editor decline
- Larry Sanger — the co-founder turned critic
References
- Giles, Jim. "Internet encyclopaedias go head to head" — Nature 438, 900–901 (2005); Britannica's rebuttal and Nature's response (2006).
- Mesgari, Okoli, Mehdi, Nielsen, Lanamäki. "The sum of all human knowledge: A systematic review of scholarly research on the content of Wikipedia" — JASIST 66(2) (2014).
- Reavley et al. "Quality of information sources about mental disorders: a comparison of Wikipedia with centrally controlled web and printed sources" — Psychological Medicine (2012).
- Randall, Eric. "How a Raccoon Became an Aardvark" — The New Yorker (2014), the citogenesis case study.
- Silva, Marco. "Alan MacMasters: How the great online toaster hoax was exposed" — BBC News (2022).
- Seigenthaler, John. "A false Wikipedia 'biography'" — USA Today (2005).
- Wikipedia, "Reliability of Wikipedia" — the project's own meta-article; primary source material.
- Wikimedia Foundation research on editor distribution, article development levels, and BLP policy history.
Footnotes
- ↑ Giles, "Internet encyclopaedias go head to head," Nature 438 (2005)
- ↑ Criticism of Wikipedia; List of Wikipedia controversies
- ↑ Randall, The New Yorker (2014)
- ↑ Ideological bias on Wikipedia
- ↑ Criticism of Wikipedia