Wikipedia’s Bot Traffic: Automated vs Human Pageviews

Home › Guides › Wikipedia’s Bot Traffic: Automated vs Human Pageviews

✓ checked 11 October 2026 · sources: Wikimedia Analytics pageviews API (data CC0), Wikimedia Foundation posts · method · caveats · download the data

Wikimedia counts every pageview as coming from a person (“user”), a self-identified crawler (“spider”) or traffic its systems judge to be automated. Since January 2023, the “automated” count has grown about 3.2 times across all Wikimedia projects while the “user” count has fallen. In September 2026, automated pageviews were 55.5% as large as user pageviews.

Read this first. These are Wikimedia’s own classifications, and they change. Wikimedia says some automated traffic has “evaded our bot detection methods” and been counted as users, and it corrected its data in 2025. We use the figures as published on 11 October 2026. Month-to-month swings are large, so look at the trend, not any single month.

automated pageviews, Sep 2026 (all projects) 7.17Bas a share of user pageviews 55.5%user pageviews since Jan 2023 -22.3%

Headline numbers

1. All projects: automated pageviews rose from 2.24B in January 2023 to 7.17B in September 2026. Over the same period user pageviews went from 16.63B to 12.92B, a change of -22.3%. Automated pageviews equalled 13.5% of user pageviews in January 2023 and 55.5% in September 2026.
2. The highest automated month so far is Dec 2025: 7.25B automated pageviews, 53.0% of the user count. September 2026 has the highest ratio of any month since January 2023. In 8 of the 45 months since January 2023, automated pageviews were more than 40% of user pageviews.
3. English Wikipedia alone: automated pageviews went from 0.98B to 1.70B (about 1.7 times), user pageviews from 7.91B to 6.27B (-20.7%). The automated-to-user ratio rose from 12.4% to 27.2%.
4. Adding self-identified crawlers (“spiders”), bots make up 50.7% of all pageviews across projects in September 2026, up from 30.9% in January 2023.
5. Year on year, automated pageviews across projects are up 18% (September 2026 against September 2025), though September 2025 was itself a high month.

Automated pageviews relative to user pageviews

Selected months

What Wikimedia says

On 5 October 2026 the Wikimedia Foundation reported “excessive data downloading” by agents it believes OpenAI operates: “millions of automated requests to our public APIs”. It added that, in 2025, “65% of the most resource-consuming traffic on its projects was coming from bots”. We have not checked those statements, and the post is Wikimedia’s account, not a finding by a court or regulator.

In March 2026 Wikimedia wrote that it was “currently blocking or throttling about 30%” of automated requests from crawlers that don’t follow its policies. Its February 2026 metrics report warns: “Since 2024, we have received progressively increasing amounts of bot traffic.” Wikimedia’s data-issues note says the 2025 correction used a new detection heuristic deployed on “August 28, 2025”, with historical data backfilled by 8 October 2025, so months from March 2025 onward were recalculated. Earlier months were not.

For what sites say about AI crawlers in robots.txt, see our AI opt-out survey and the AI-Crawler & llms.txt Census. For the court cases about scraping, see how scraping cases end.

Method

Data. Wikimedia Analytics pageviews API, aggregate endpoint, all access methods, monthly, for all-projects and en.wikipedia and each agent type (user, automated, spider), January 2020 to September 2026, fetched 11 October 2026. The “automated” series starts in April 2020; we use January 2023 onward.

Calculations. Ratios divide the automated count by the user count for the same month and project. “Bots” adds automated and spider counts. Percentage changes compare January 2023 with September 2026 unless stated.

Verification. A separate script re-reads the saved API responses, recomputes every figure and checks each quote against the saved Wikimedia pages.

Caveats

• “Automated” and “user” are Wikimedia’s classifications. Some bots are still counted as users, so the true human decline is probably larger, as Wikimedia itself says.
• Classification methods changed during the period (see above), which can move counts without any change in behaviour.
• Pageviews count requests for pages. They don’t show which company is behind the traffic, and they don’t include API or media downloads.
• The monthly figures are volatile. A rise in one month isn’t a trend.
• Wikimedia’s claims about specific operators are Wikimedia’s statements; we haven’t verified them.

Download the data

• Wikipedia pageviews by agent type (CSV): monthly user, automated and spider pageviews for all projects and English Wikipedia, January 2023 to September 2026, with the automated-to-user ratio.

Cite as: ProxyPicker, “Wikipedia’s Bot Traffic: Automated vs Human Pageviews”, 11 October 2026, https://proxypicker.com/wikipedia-bot-traffic/. Data: Wikimedia Foundation, pageviews API (CC0).

Related: AI-Crawler & llms.txt Census · The AI Opt-Out Standard vs Real robots.txt Files · How Scraping Cases End · AI Chatbots in Chrome Traffic Rankings.

Press & data requests: email [email protected] for interviews, the raw data behind this page, or corrections. See also our contact page.