Newshot

The New Era of Wikimedia Traffic: Thinking Beyond Pageviews

Original text(via Lobsters)

Part 1 of Wikimedia in a Changing Traffic Ecosystem: This post begins a series for Wikimedia communities, and other interested readers, exploring some recent readership trends and what they mean for the movement’s future. Drawing on research from across the Wikimedia Foundation, we’ll examine the forces shaping our traffic, the implications for readers and contributors, and how these findings can inform our strategy.

For more than two decades, billions of visits to Wikipedia and other Wikimedia projects have reflected something remarkable: people around the world choose to seek out, share, and contribute to free, human knowledge. Pageviews have long been one of Wikimedia’s most important measures of readership, helping us understand how people access and engage with our projects.

But the internet around Wikimedia is changing, and pageviews are becoming a less reliable measure of human readership and demand for Wikimedia knowledge. While Wikimedia content is more valuable than ever on the internet, it is becoming less visible. Search summaries, AI assistants, and social platforms increasingly provide information from Wikimedia without requiring people to visit Wikimedia projects. This helps share Wikimedia’s knowledge with more people, but it also means fewer people visit our websites and get involved as an editor or donor. At the same time, as automated systems access Wikimedia content at unprecedented scale, it is becoming harder to understand how, where, and whether that information ultimately reaches people.

Over the past year, the Wikimedia Foundation has been investigating this changing traffic ecosystem and reader behaviour through a broader body of research. In this first installment of a new research series, we share what we’ve learned and introduce a framework for understanding Wikimedia’s traffic health beyond pageview counts alone. In the posts that follow, we’ll take a closer look at what our research reveals about how people discover, read, and contribute to Wikimedia projects, sharing more detailed data and findings across traffic sources, devices, and Wikimedia projects.

What we’ve learned since our last traffic update

In October 2025, we published New User Trends on Wikipedia, highlighting emerging challenges in readership and contributor growth.

Since then, our research has continued to examine changes in human pageviews over a longer period of time to understand longer-lasting patterns. Our findings show that human readership is declining, particularly as fewer readers arrive through search engines. However, changes in how we identify automated traffic make it difficult to determine the precise extent of this decline. To better understand these changes and the measurement challenges that complicate their interpretation, our research identified three broad periods in Wikimedia’s pageview patterns since 2020:

  • A pandemic-era readership surge (2020–2021), when global information-seeking behavior shifted substantially and people were spending more time at home, driving increased readership and unusually strong account creation rates. For some, Wikipedia also became a vital source of timely, reliable COVID-19 information, with volunteer-maintained content reaching millions of readers and being reused by other platforms. Mobile web also continued to gain ground relative to desktop readership during this period.
  • A period of changing traffic patterns and measurement challenges (2022–Early 2025), as search engines continued expanding their direct-answer features, short-form video and social platforms became important channels for discovering information, and automated pageviews complicated our understanding of readership. Some automated activity inflated reported pageviews, while subsequent measurement corrections improved our ability to distinguish human from automated pageviews.
  • A period of declining measurable human pageviews (Early 2025–Present), particularly from declining search engine referrals beginning in 2025, alongside continuously increasing automated pageviews. At the same time, a growing share of likely human pageviews arrives without identifiable referral information: while some of this may reflect readers discovering Wikimedia through platforms that don’t always share referral information, or through privacy tools that obscure referral information, other pageviews may be automated, making it difficult to determine how discovery patterns are changing.

Fig. 1. Monthly Human Pageviews Highlighting Traffic Eras since 2020

Our analysis shows that Wikimedia is currently experiencing longer-lasting changes in how people arrive at and interact with our projects. This is in line with findings from the Wikimedia Foundation’s Brand Health Tracker, which show declining awareness and visibility of Wikipedia from 2025 onwards, particularly through search, even as trust in Wikipedia remains relatively stable. We also hear the user perspective on these trends in recent user experience research: that users themselves cite search summaries and AI chatbots as the key reasons why they don’t visit Wikipedia more often, and use it less than they used to.

We are sharing our findings, based on publicly available data, to illustrate these trends and explain what we are learning about changes in the internet. While the data shows that human readership is declining, for reasons we describe in this post, we cannot provide a precise estimate of the decline. Changes in automated traffic and how we detect it make it difficult to distinguish changes in actual readership from changes in measurement over time. In the next section, we explain this challenge.

Why measuring traffic has become complicated

One of the biggest challenges in measuring readership is distinguishing pageviews generated by people from those generated by automated systems.

Automated systems, including search crawlers, AI training scrapers, AI grounding services, and other bots, access Wikimedia content for many reasons. Some of this activity supports the wider distribution of free knowledge. But most of it places substantial demands on our infrastructure that we increasingly need to address – while Wikimedia content is free, its infrastructure is not.

It is an ongoing challenge to identify which pageviews are automated. Many of the proprietors of bots and scrapers are constantly evolving their approach so that their pageviews appear human. That means that we need to constantly evolve our approach to accurately categorize these pageviews. We made large improvements to bot detection in March 2025 and again in late June 2026, including the introduction of KAPOW, a new system for this purpose. These improvements have lowered reported human pageview counts by reclassifying traffic previously identified as human as automated traffic. This effect causes our reading of human pageviews to be clouded – they are a “moving target”.

An important limitation when interpreting historical pageview trends stems from Wikimedia’s longstanding practice of protecting the privacy of our users, which includes discarding detailed activity logs on a regular basis. This prevents us from applying the new KAPOW automated traffic detection methodology to data from more than 90 days in the past. As a result, we cannot consistently distinguish between changes in measurement and changes in user behavior over longer periods, making direct year-over-year comparisons difficult to interpret. For this reason, we are not providing specific year-over-year decline figures.

Another challenge is understanding how readers discover Wikimedia beyond traditional search engines. As people increasingly encounter and share information across social platforms, messaging apps, and other online spaces, many of these channels do not consistently pass along referral information. Missing referral information can also result from direct visits, browser privacy protections, and automated pageviews, making it difficult to distinguish between different pathways to Wikimedia. Greater transparency and stronger attribution from websites and apps, including consistently sharing referral information, would help us better understand where our readers come from.

Looking beyond pageviews: A framework for traffic health

Understanding Wikimedia’s traffic health requires looking beyond overall pageview totals to examine where changes are occurring and what they mean for readers, contributors, and the broader ecosystem.

A decline can mean very different things depending on where it occurs. Is it concentrated among readers arriving through search? Is it affecting mobile web more than other platforms? Are some language editions of Wikipedia experiencing greater disruption? And what does it mean for the people who might eventually become contributors and/or donors?

To investigate these questions, we developed an exploratory Traffic Health framework, which considers four interconnected dimensions:

  • Visibility: what do estimated human pageviews tell us about Wikimedia’s reach?
  • Visitor Outcomes: to what extent do pageviews lead to meaningful outcomes (e.g., longer reading time, account creation, editing, donation)?
  • Ecosystem resilience: how dependent is Wikimedia on particular traffic sources or platforms, and how vulnerable are different parts of the ecosystem to disruption?
  • Structural change and characteristics: are we seeing temporary fluctuations, or have pageviews entered a different long-term pattern?

We examine these questions across device types, referral sources, wiki types, and individual Wikimedia projects.

This approach has already revealed differences that are difficult to see in topline pageviews. For example, the majority of human readers access Wikipedia through mobile web, which is particularly vulnerable to changes in search engines as a higher proportion of mobile web pageviews come from search referrals as a source. In contrast, Wikipedia’s mobile app has comparatively stable pageview patterns as it relies less on search engine referral traffic. The relative stability of app pageviews highlights an important opportunity to strengthen direct relationships with readers.

We are still evolving this framework. Some of its indicators are exploratory, and they are intended to help identify important changes rather than establish their causes. Together, they are a way for us to identify vulnerabilities, recognize areas of resilience, and find opportunities to strengthen Wikimedia’s reach and engagement. This framework will also guide the research shared in subsequent posts, as we explore different aspects of traffic health across sources, devices, and Wikimedia projects.

FIGURE 2: Visualization of Traffic Health Framework

Meeting this moment

These findings have implications beyond how we report pageviews. The Wikimedia Foundation’s 2026–2027 Annual Plan contains several efforts to adapt to the changing internet, build on Wikimedia’s strengths, and strengthen its long-term sustainability:

  • Increasing visibility in traditional and AI search: Improving how Wikimedia content is discovered and indexed by search engines and AI-powered search experiences to help readers continue finding our projects.
  • Being present in more places: Exploring new distribution opportunities and partnerships, bringing Wikipedia to where people are, putting our content into places where people already spend time, creating new visibility beyond search. Improving attribution so that wherever people encounter Wikimedia knowledge they can recognize its source and find their way to us.
  • Direct connections with readers: Encouraging readers to create accounts so that they can more deeply explore Wikipedia and get closer to becoming contributors, encouraging them to use the Wikipedia mobile apps as a go-to place for learning, and providing new readers with the types of experiences that make them want to return to learn from Wikipedia.
  • Making it easy to be a contributor: Making it compelling for readers to try contributing, making it increasingly easy to edit from mobile devices, and supporting advanced contributors with the tools they need to run the wikis at a higher scale.
  • Protecting our infrastructure: Addressing abusive bot pageviews and encouraging responsible content reuse through appropriate APIs and services, including Wikimedia Enterprise, so that we can continue to be the most widely accessible open knowledge resource.

Our ongoing research can help inform these priorities by identifying where pageview trends are changing, which audiences are most affected, the impact of our efforts to respond to these trends and what additional evidence is needed to evaluate potential responses.

It also reflects a broader commitment to transparency. In a rapidly changing information ecosystem, we believe it is important to share not only what we know, but also what we are still trying to understand.

What comes next and how you can contribute

This is the first installment in a series exploring how Wikimedia’s traffic ecosystem is changing.

In the coming posts, we’ll explore additional topics in greater depth:

  • Search and readership: How declining search referrals affect Wikimedia’s visibility and readership patterns.
  • Beyond search: The changing role of alternative pageview sources, automated pageviews, and the challenge of understanding where readers come from.
  • Mobile readership: Why Mobile Web and apps face different challenges and opportunities for reach, engagement, and resilience.
  • Readers and contributors: How changes in readership relate to Wikimedia movement participation.
  • Different Wikimedia projects: Why pageview trends and their implications vary across projects and communities.

We also want to learn from the people who know these projects best: the communities who create, maintain, and improve Wikimedia knowledge every day.

Have you noticed changes in how people find or use your Wikimedia project? Are there local trends, research findings, or questions you think we should investigate? We welcome your perspectives as we continue this work. Please join the conversation on Meta-wiki.

Understanding Wikimedia’s future requires more than counting visits. It requires understanding how people discover knowledge, what brings them back, and how they become part of the communities that make that knowledge possible. With a clearer picture of these changes, we can make better decisions about how to reach and serve readers in the years ahead.

Share this:

  • Share on Mastodon (Opens in new window) Mastodon
  • Share on Bluesky (Opens in new window) Bluesky

Can you help us translate this article?

In order for this article to reach as many people as possible we would like your help. Can you translate this article to get the message out?

Related

The text above was extracted automatically from the source page and may miss images, tables or code blocks. The original is authoritative.

Published by Lobsters
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗