Wikipedia reports traffic drop, blames AI search and social media

Wikipedia has reported a reduction in readership, blaming the dip on rise in search engines’ AI summaries, bot traffic and a new information environment.

Between March and August this year, human pageviews decreased by 8% compared with the same period in 2024. At the same time aggressive bots and crawlers have been bombarding the online encyclopaedia, putting a strain on its infrastructure.

Like most websites, Wikipedia tracks visitor numbers and has systems to differentiate between human users and bots so the latter can be policed. The site has been a prime source of training data for AI models such as GPT and Claude, and in their hunger for fresh data the scraper bots have become sneakier, emulating human visitors to avoid being classified as automatons.

In a blog post Marshall Miller of Wikipedia parent the Wikimedia Foundation describes how a spike in traffic from Brazil in May turned out to be bots designed to appear human. On revising its detection logic, genuine visits, which had appeared to be holding up in the face of changes wrought by GenAI, were in fact found to be down by about 8%.

Detecting bots has become increasingly expensive and time consuming for an organisations that runs largely on public donations and voluntary labour. Meanwhile, the AI models trained on the data are being used in search engine summaries, and Miller says that fewer people are now clicking through to the site.

“Search engines are increasingly using generative AI to provide answers directly to searchers rather than linking to sites like ours.”.

Wikipedia is not alone in experiencing falling pageviews as a result of AI search summaries. A study in March found a huge reduction in referral traffic to 160 different sites from AI engines compared to traditional Google search, together with a massive increase in content scraping.

Wikipedia and other publishers are also facing pressure from a change in the information landscape. Younger users increasingly use video platforms such as TikTok and YouTube as primary sources rather than traditional websites.

But with Wikipedia being one of the primary feedstocks for LLMs, a decline in real traffic may lead to fewer donations and voluntary activity, which could ultimately lead to a vicious circle where training data and model outputs becomes less and less reliable.

To counter the threat, Wikimedia is developing policies and technical capabilities to ensure that third-party platforms properly credit and link back to Wikipedia.

There is also an effort to understand how readers discover and use the online encyclopaedia, a drive to recruit new volunteers, and experiments with new ways to reach younger audiences whose main information sources are social video and gaming platforms.

“25 years since its creation, Wikipedia’s human knowledge is more valuable to the world than ever before. Our vision is for a future where everyone can participate in knowledge creation and sharing – a future that is possible when everyone uses the free knowledge ecosystem responsibly,” Marshall writes.