UA

Wikipedia in 2026: Key Statistics

A reference sheet of Wikipedia's key numbers for 2026 — scale, traffic, editors, the AI relationship and the paid-editing market. Every figure below carries its date and comes from a primary source: Wikimedia Foundation publications, live Wikimedia APIs, or the community's own records. Cite freely; the dates are part of the data.

Scale: articles and languages

  • 7,256,820 articles in English Wikipedia (live API, 2026-10-29) — within 66.5 million total pages including redirects and service pages, two numbers that must not be confused.
  • 65+ million articles across 300+ languages per the Wikimedia Foundation's 25th-anniversary figures (2026-01-15); the Foundation's October 2026 self-description raises it to more than 67 million.
  • 348 active language editions — 364 Wikipedias officially created, 16 closed (Meta, 2026-10-29).
  • Wikidata: ~123.8 million items (live API, 2026-10-29) — the structured-data layer that feeds knowledge panels and AI systems.

Takeaway: "Wikipedia" is 348 separate editorial communities around one brand — which is why notability, sourcing and pricing all run per edition.

Traffic: still huge, visibly shifting to AI

  • 12.92 billion human page views across all Wikimedia projects in September 2026 — down 7.9% year over year (Wikimedia REST API); English Wikipedia alone took 6.27 billion (−9.3% y/y).
  • 1.66 billion unique devices accessed Wikipedia sites in September 2026 (−7.5% y/y); the July 2026 seasonal peak was 2.03 billion.
  • The Foundation's own assessment (2025-10-17): human pageviews are down roughly 8% year over year, which it attributes to generative AI and social media answering questions directly, often based on Wikipedia content.
  • Independent context: when a Google AI summary appears on a results page, users click a traditional link in 8% of visits versus 15% without one (Pew Research, March 2025 data).

Takeaway: people increasingly read Wikipedia through AI intermediaries — the content's influence persists even as direct visits decline, which changes where accuracy matters.

Editors: a stable core, a shrinking periphery

  • Nearly 250,000 editors per month across all languages — the Foundation's definition: at least one edit per month, named accounts only (WMF, 2026-01-15). Wikipedia is edited 324 times per minute.
  • 40,940 active editors on English Wikipedia in September 2026 (5+ content edits in the month) — down 14.5% year over year.
  • 4,284 very active editors (100+ edits/month) — down just 1.7%: the committed core holds while casual participation thins.
  • Different public "editor counts" use incompatible definitions (≥1 edit vs. ≥5 edits vs. any action in 30 days) — compare only like with like.

Takeaway: fewer, more concentrated editors review more content — review queues and scrutiny per article both grow, which raises the bar for anything submitted.

AI and Wikipedia: the feedback loop

  • The Foundation states that almost all large language models train on Wikipedia datasets and calls Wikipedia one of the highest-quality datasets used in LLM training (WMF, 2025-10/2026-01).
  • In a 730,000-conversation study of ChatGPT usage (US, Oct–Dec 2025), Wikipedia was the single most-cited domain: 5% of all citations, appearing in 18% of conversations that cite sources.
  • Bots now dominate the expensive end of traffic: at least 65% of the costliest traffic comes from bots while bots are ~35% of pageviews (WMF, April 2025); by March–April 2026 the Foundation was blocking or throttling ~30% of crawler requests that ignore its policies.
  • Meanwhile the llms.txt standard remains unconsumed in practice — 97% of llms.txt files never receive a single request (137K-site study, May 2026), and Google stated (2026-07-10) its AI features need no special files or markup.

Takeaway: AI assistants read Wikipedia at industrial scale and cite it more than any other domain — being accurately represented there is now machine-distribution infrastructure, not vanity. Context: the industry page and AI visibility.

The paid-editing market, in the community's own numbers

  • The community-maintained list of paid-editing companies recorded, as of 2026-10-05: 197 entries under "Undisclosed paid editing", 11 under "Disclosed paid editing", 2 disclosed-but-blocked, and 19 defunct.
  • Review is slow and selective: a Signpost recap (2026-05-22) of an industry audit of 1,009 Articles-for-Creation drafts from late 2025 reported 68% declined, with notability, unreliable sourcing and AI-written text as the top decline reasons, and average review times around 30.5 days for accepted drafts. (The Signpost itself flags the audit as non-peer-reviewed; treat it as directional.)
  • The disclosure rules behind those numbers are not new: the Wikimedia Terms of Use have required paid-contribution disclosure since 16 June 2014, with the current ToU in force since 7 June 2023.

Takeaway: the community classifies the overwhelming majority of known paid-editing vendors as operating undisclosed — which is why disclosed status is a trust signal in itself, and why vendor vetting is not optional.

The alternative encyclopedias

  • Grokipedia — the AI-generated encyclopedia launched on 2025-10-27 — stood at version 0.3 as of 2026-10-29, with its article count self-described in 2026 as over 6 million (sitemap evidence closer to 5.9 million). Its editorial model and review practices remain materially different from Wikipedia's volunteer review.
  • Wikipedia remains the only nonprofit site in the global top-10 most visited (WMF, 2026-01-15) — the trust position the alternatives are competing against.

Takeaway: alternative wikis can complement a presence strategy, but none currently substitutes for Wikipedia's review-earned authority with either humans or AI systems. More terms: glossary.

Frequently asked

How many articles does Wikipedia have in 2026?

English Wikipedia held 7,256,820 articles as of 2026-10-29 (live API), and the Wikimedia Foundation describes the project as more than 67 million articles across over 300 languages as of October 2026. The larger "total pages" figure (~66.5 million for English alone) includes redirects and service pages and should not be quoted as an article count.

Is Wikipedia's traffic declining?

Human page views are declining moderately: 12.92 billion across all projects in September 2026, down 7.9% year over year, with the Foundation's own estimate at roughly −8% and the stated cause being generative AI and social media answering questions directly. Total influence is a different story — AI assistants cite Wikipedia more than any other domain, so its content increasingly reaches people through intermediaries rather than visits.

How big is the paid Wikipedia editing market?

The community's own registry gives the clearest structural number: 197 firms listed under undisclosed paid editing versus 11 under disclosed, plus 2 disclosed-but-blocked and 19 defunct, as of 2026-10-05. No reliable revenue total exists for the market; what the numbers show is its compliance structure — the overwhelming majority of known vendors operate outside the disclosure rules.

How hard is it to get a new article accepted through Articles for Creation?

Directionally hard: a Signpost recap (2026-05-22) of an industry audit covering 1,009 AfC drafts from late 2025 reported 68% declined, with notability, unreliable sourcing and AI-written text as the leading decline reasons, and roughly 30.5 days average review time for accepted drafts. The Signpost flags the audit as non-peer-reviewed, but the direction matches the shrinking-reviewer, rising-scrutiny environment the editor statistics show.

Continue reading