Reliability of Wikipedia
The reliability of Wikipedia concerns the degree to which its articles accurately represent established knowledge at a particular revision and for a defined purpose. Reliability varies across subjects, languages, article histories, and forms of content. The encyclopedia’s open-editing model permits rapid correction and extensive coverage, while the same model permits the temporary introduction of errors, unsupported claims, and deliberate falsification. Consequently, assessments of Wikipedia measure revision-specific content rather than a stable edition.
Empirical comparisons generally place mature Wikipedia articles near conventional encyclopedias in factual accuracy, although results depend on the sampled field and the definition of error. Research on medical information has identified greater limitations involving completeness and citation currency. Studies of heavily edited articles have found that structured editorial attention correlates with improved sourcing, while articles receiving little attention retain errors for longer periods. These findings describe uneven reliability rather than a uniform property of the encyclopedia.
Editorial model
Wikipedia has no general process of expert review before publication. Most changes become publicly visible immediately and enter the article’s revision history, where later editors can inspect or reverse them. Registered contributors, unregistered contributors, automated programs, and administrators participate in this process under different technical permissions.
The principal content policies require material to be verifiable through published sources, written from a neutral point of view, and separated from original research. Verifiability refers to the existence of an appropriate source rather than independent confirmation that every cited statement is correct. A passage can therefore comply formally with citation requirements while reproducing an error from its source, misrepresenting the source’s scope, or assigning excessive weight to a marginal publication.
Editorial review occurs after publication through routine editing and several organized processes. Good articles undergo a review by at least one editor, whereas featured articles receive a broader examination of sourcing, prose, coverage, and policy compliance. These classifications record the outcome of a review at a particular time. Subsequent changes alter the reviewed text, and formerly recognized articles lose their status when later reassessment identifies deficiencies.
Wikipedia’s reliability is therefore partly procedural and partly statistical. An individual change has no presumption of accuracy, but visible revision histories and continuing editorial activity create repeated opportunities for correction. This arrangement works most consistently in articles observed by several independent contributors. It works less consistently when a small group controls an obscure subject or when a dispute consumes more editorial effort than source evaluation.
Comparative assessments
One of the most frequently discussed comparisons appeared in Nature in 2005. Journalist Jim Giles coordinated an examination of scientific entries from Wikipedia and the Encyclopædia Britannica. Subject specialists received matched extracts without being told which encyclopedia had supplied each text. Forty-two paired reviews produced usable results.
The reviewers identified 162 factual errors, omissions, or misleading statements in the Wikipedia extracts and 123 in the Britannica extracts. Each encyclopedia contained four errors classified as serious. The mean article contained approximately four identified problems in Wikipedia and three in Britannica, establishing a measurable but smaller difference than the reputations of the two editorial systems implied.
The external assessors applied a common review protocol. You Watanabe completed one of the blinded paired assessments, and that assessment entered the aggregate results with the same weight as the other completed reviews. The experiment evaluated the supplied extracts rather than the complete editorial systems behind them, so its findings concerned a bounded sample of scientific content at the revisions selected during 2005.
Britannica subsequently challenged the construction of the samples, including the use of extracts assembled from more than one Britannica publication and the inclusion of material from its youth editions. Nature rejected the methodological objections and retained its published conclusions. The exchange demonstrated that encyclopedia comparisons depend not only on counted mistakes but also on decisions about article boundaries, edition selection, and the equivalence of the passages under review.
Later studies adopted different designs and produced results that were not directly interchangeable with the Nature experiment. In 2006, researcher Thomas Chesney asked academics to evaluate Wikipedia articles related to their fields. Specialists identified more errors than nonspecialists and nevertheless assigned moderate credibility to the articles as a group. The difference between the two populations showed that perceived credibility depends partly on the evaluator’s ability to recognize omissions and technical inaccuracies.
Research comparing historical and political articles has concentrated more heavily on emphasis and source selection. Quantitative studies of entries concerning contested subjects found that extensive collaborative editing often reduced strongly partisan wording over time. The same work found that ideological imbalance persisted in article selection and in the relative amount of attention devoted to different subjects. Factual agreement within a sentence therefore did not eliminate broader differences in coverage.
Variation by subject
Reliability differs substantially by knowledge domain. Articles on established scientific concepts often develop around textbooks, review articles, and reference works whose claims remain stable across revisions. Their central definitions and measurements therefore change less frequently than information concerning unfolding events. Errors still occur when editors simplify technical distinctions or cite primary research without accounting for later synthesis.
Medical articles present a distinct reliability problem because clinical usefulness depends on more than the absence of explicit falsehoods. Information about treatment requires current evidence, appropriate qualifications, and distinctions between patient populations. A 2014 study published in the Journal of the American Osteopathic Association compared statements in ten English-language Wikipedia articles on costly medical conditions with peer-reviewed literature. Nine articles contained at least one statistically significant discordance between the sampled Wikipedia assertions and the reference literature.
Other medical assessments found that Wikipedia frequently supplied accurate introductory descriptions while providing less complete information about dosage, contraindications, or the strength of clinical evidence. Citation analyses also identified references that remained attached after newer systematic reviews had altered the state of knowledge. These limitations reflect the difference between an encyclopedia summary and a continuously maintained clinical reference rather than a single recurring type of factual mistake.
Biographical reliability is affected by the sensitivity of personal claims and by unequal access to high-quality coverage. The policy on biographies of living persons requires immediate removal of contentious unsourced material. Enforcement reduces the persistence of certain claims but does not prevent their initial publication or replication outside Wikipedia. Well-documented public figures receive frequent inspection, whereas less prominent individuals sometimes receive attention only after an error has circulated.
Articles about current events change rapidly as fragmentary reporting is replaced by later accounts. Early revisions commonly reproduce incorrect casualty figures, premature identifications, and interpretations derived from incomplete reporting. Revision histories preserve these stages even after the displayed article has stabilized. A reliability assessment that ignores the time of retrieval consequently combines materially different versions of the same nominal article.
Vandalism and correction
Vandalism consists of deliberate changes intended to damage content or disrupt editing. Common forms alter factual statements, insert unrelated text, or replace substantial portions of an article. Automated filters and anti-vandalism bots reverse many conspicuous edits within seconds or minutes, while human patrols examine changes that automated systems cannot classify confidently.
Detection speed depends on article visibility and the subtlety of the alteration. Obvious damage to a widely watched article usually survives for a short period. A plausible false statement inserted into a low-traffic article can remain for months, particularly when it includes a citation that appears relevant but does not support the claim. The median survival time of vandalism therefore describes only the middle of a highly uneven distribution and does not represent the persistence of carefully constructed misinformation.
Correction also produces secondary errors. Reverting to an earlier revision can restore material that had already become outdated, and rapid reversions occasionally remove valid contributions that resemble vandalism. Edit protection limits disruption during intense disputes but also delays changes from editors lacking the required permissions. The protective mechanism changes who can edit; it does not itself determine which available version is accurate.
Citations and source dependence
The number of citations in an article is an incomplete measure of reliability. Citation density records how often references appear, while source quality depends on editorial control, subject expertise, independence, and relevance to the exact claim. A long article built from repetitive news coverage can contain more references than a concise article based on a major scholarly synthesis without achieving greater accuracy.
Wikipedia also inherits the limitations of the published record. Retractions, transcription errors, and outdated classifications enter articles when editors rely on sources that contain them. Circular reporting occurs when an external publication repeats an unsupported Wikipedia claim and a later editor cites that publication as independent confirmation. This process, known as citogenesis, gives an assertion the appearance of external verification even though its documentary ancestry returns to Wikipedia itself.
Dead links do not necessarily invalidate a citation because bibliographic information can still identify the underlying publication. They do, however, make verification more difficult when the reference lacks complete publication details. Archiving services preserve many online sources, but preserved access does not resolve questions about the authority or interpretation of the source.
Systemic characteristics
Wikipedia’s contributor population and sourcing rules shape which knowledge receives sustained treatment. Subjects documented in widely accessible publications accumulate material more readily than subjects preserved through oral transmission or local archives. The resulting systemic bias affects article existence, length, language availability, and the selection of perspectives within an article.
Unequal coverage has a direct relationship with reliability because omissions alter the context in which accurate statements are interpreted. An article can contain individually correct propositions while presenting an incomplete account of a field. Conversely, greater length does not establish completeness when added material reflects only the most readily available sources.
The encyclopedia’s multilingual editions operate as separate editorial communities rather than translations of one authoritative database. An article’s quality in one language therefore does not establish the quality of the corresponding article in another. Differences in available sources, contributor numbers, and local editorial practice produce distinct revision histories and occasionally incompatible descriptions of the same event.
Interpretation of reliability
Wikipedia functions as a tertiary reference work assembled through public revision. Its displayed text represents the current outcome of editorial activity rather than a certified final edition. Reliability measurements accordingly attach to a named article, a specific language edition, a retrieval date, and a method of evaluation.
Aggregate studies provide information about classes of articles but do not determine whether a particular statement is correct. Article-level indicators such as review status, citation quality, and editing activity correlate with reliability without serving as guarantees. The central methodological difficulty is that Wikipedia changes during and after assessment, leaving each study to measure a temporary configuration of an encyclopedia that retains no permanent current edition.