Research Report Nº 13
The thirteenth in a series of biannual reports from Wikimedia Research, published every June and December.
Foreword
Welcome to our Research Report No. 13. As we close out 2025 and look toward the horizon, this edition carries a special significance. In less than a month, on January 15, 2026, Wikipedia will celebrate its 25th birthday. Reaching a quarter-century of free, volunteer-led knowledge is an extraordinary milestone—one that reminds us that Wikipedia is not a collection of data, but "humanity at its best".
Reflecting on the past year, I am filled with immense gratitude for our team of research scientists, research engineers, user experience and design researchers, survey specialists, analysts, and community specialists as well as our formal collaborators, research fellow, contractors, interns, and other volunteers who worked with us. Their dedication has been the backbone of the work detailed in these pages. Your dedication, diverse perspectives and rigorous inquiry ensure that our movement remains resilient in a rapidly changing information landscape.
In this report, you will see the results of our increased investments in understanding readership and search. We have completed deep-dive research into reader discovery needs, started research on Wikipedia reader use cases and user needs, and launched a large-scale global survey across 11 languages. This work is critical as we navigate an era of further intermediation, where ensuring readers can find and engage directly with neutral, high-quality content is more important than ever.
You will also find updates on our human-centered AI strategy. At a time when the internet is increasingly awash in synthetic content, our commitment is clear: we are investing in AI that supports, rather than replaces, human agency. Our focus remains on conducting research that can help alleviate the "mundane" tasks for editors—such as vandalism detection and information retrieval—so they can focus on the nuanced work of deliberation, judgment, and consensus-building that defines our projects.
Finally, I want to draw your attention to the "trends to watch" section at the end of this report. I have shared some reflections there on the evolving challenges of knowledge integrity and the future of Neutral Point of View (NPOV) in an AI-driven world. I invite you to read that section closely for a deeper look at the questions we are facing and we hope to work with you on as we enter Wikipedia’s next chapter.
Thank you for being part of this journey. Here is to 25 years of free encyclopedic knowledge, and to the many more to come.
– Leila Zia, Director and Head of Research at the Wikimedia Foundation
Projects
The large majority of the work of our team is done in the open and reported here. We also provide consultation to other teams at the Wikimedia Foundation (WMF) or external researchers and not all of that work has public outputs. We strive to share what we can and will include these additional projects as they become publicly available in future Research Reports.
-
Address Knowledge Gaps
We develop models and insights using scientific methods to identify, measure, and bridge Wikimedia’s knowledge gaps.Reader Survey 2025
We fielded the global reader survey across 11 Wikipedia languages and started the data processing. (Learn more)
Discovery needs of Wikipedia readers
We completed research into understanding the needs and opportunities for improving discovery and search for readers. (Learn more)
Wikipedia reader use cases and user needs
We conducted qualitative and quantitative research and started documenting and communicating the results. (Learn more)
Semantic search prototype
We developed the logic for a prototype built by the ML team that accepts natural language questions and returns paragraphs from relevant articles. (Task)
Semantic search prototype evaluation
We conducted a qualitative evaluation of the semantic search prototype compared to the existing keyword search. (Learn more)
Search literature review
We conducted a review of literature on how readers use on- and off-wiki search and provided recommendations to improve search. (Learn more)
Search evaluation dataset
We started to develop a benchmark dataset to evaluate and improve search on Wikipedia. (Task)
LLMs for text simplification
We concluded this project. (Learn more)
Newcomers survey
We conducted a survey of new editors on English Wikipedia to understand who they are and how we can improve their experience. (Learn more)
AI for mentorship
We started exploring what types of questions are asked by new editors on-wiki and what challenges arise in answering those questions. (Learn more)
Guided article and section creation
We investigated how editors create new articles and how reviewers evaluate them as part of our efforts to test guided creation concepts. (Learn more)
-
Improve Knowledge Integrity
We develop models and insights using scientific methods to support the technology and policy needs of the Wikimedia projects in the areas of misinformation, disinformation, and content integrity.Admin trends and potential causes
We submitted the research for peer-review to a research conference and held multiple sessions during Wikimania 2026 to raise awareness about the results. (Task)
A model for moderation detection
We did not conduct additional research on this front. We will resume reporting when the work resumes.
A model for Wikidata vandalism detection
We released the model binaries and ran extensive tests to support the ML team productionizing the Wikidata Revert Risk model. (Task)
Research on NPOV
We drafted a meta-method for analyzing the state of NPOV on Wikipedia. (Learn more)
Guidance on NPOV research
We published a set of recommendations for research about NPOV on Wikipedia. (Paper)
Tone Check Support
We developed a method to test the Tone Check model in languages without enough evaluation data, in support of the ML team expanding its language coverage. (Task)
Recsys for Patrollers
We built a recommendation model that provides personalized recommendations to patrollers for reviewing high risk edits and iterated based on feedback. (Task)
Template suggestions
We started research to understand the usage of maintenance templates to develop new models for suggesting article improvements to contributors. (Learn more)
Editor history prototype
We built a prototype for searching the editor history in order to increase discoverability for moderator tools such as Suggested Investigations. (Task)
-
Building the Foundations
Some of our work is in support of multiple initiatives or can inform multiple lines of future research, technology developments, and change of governance on the ground. We call this foundational work.A roadmap for AI research
We drafted a research roadmap in response to the WMF’s AI strategy for editors. (Task)
Developer Satisfaction Survey
We launched the annual Developer Satisfaction Survey on December 4th. (Learn more)
A new version of the mwedittypes library
We released a new version of the mwedittypes Python library that supports the analysis of edit diffs on Wikipedia via both wikitext and HTML. (Learn more)
A whitepaper for Wikipedia research and privacy recommendations
We revised the whitepaper based on community feedback and published it. (Paper)
AI Retraining
We developed a framework to assess model accuracy based on training with different snapshot of data, and provided recommendations for retraining frequency. (Task)
Updating research pipelines for efficiency
We implemented a library to enable research scientists to re-use production code in jupyter notebooks and to facilitate contributing new code to these repos. (Tools)
Reader direction
We started drafting an updated reader research direction to identify the most impactful open research questions in this area for the next 3-5 years. (Task)
Collection user needs
We conducted qualitative research to understand the importance of features that allo readers to save parts of articles, media form articles, or annotations. (Learn more)
Informing Edit Check
We conducted research to understand the motivations of experienced editors who develop tooling, their current usage of AI, and their reactions to Edit Check. (Learn more)
Motivating new editors
We conducted qualitative research on new editors to inform a potential on-wiki progression system. (Learn more)
Mapping non-emergency help pathways
We mapped community-created and maintained pathways to get help for non-emergency issues such as harassment to inform the design of the Incident Reporting System. (Learn more)
EU DSA policy support
We supported the Legal department in the formation of a new process for compliance with the EU DSA Article 40(4). (Learn more)
Attribution framework
We started a large-scale unmoderated study of draft attribution frameworks in support of the Design team developing a new attribution framework for web, apps, voice assistants, and LLMs. (Task)
-
Strengthening the research communities
Wikimedia projects are created and maintained by a vast network of individual contributors and organizations. We focus part of our efforts on strengthening part of this network: the Wikimedia research community.Presentations, Panels, and Keynotes
We engaged with the research and scientific communities, and supported the Wikimedia communities in advancing their work informed by research findings.
In July, we attended IC2S2 2025 (International Conference on Computational Social Science) where we engaged with members of the Wikimedia Research Community and raised awareness about Wikimedia Research.
In July, we attended ACL 2025 (Association for Computational Linguistics) where we co-organized the second WikiNLP workshop, a space to celebrate Wikimedia's contributions to the NLP community and highlight approaches to ensuring the sustainability of this relationship for years to come.
In August, we attended Wikimania 2025 in-person and virtually. We participated in Users with Extended Rights Day and raised awareness about key findings learned from research on Wikipedia administrator recruitment, retention, and attrition. We further hosted a panel with the larger Wikimania audience to raise awareness about the findings and facilitate the exchange of ideas between Wikipedia languages. Lastly, we hosted a session where we shared 5 Research Findings with the communities and facilitated conversations for the communities to reflect on the insights and how they can use them on the ground and in their work.
In September, we participated in and gave a lightning talk at the 4th annual Trust & Safety Research Conference, on the topic of conducting multilingual participatory Trust & Safety research with the wikimedia communities.
In October, we attended the 5th edition of the Mestizajes International Workshop on Science and Humanities and shared our research to assess source reliability on Wikipedia.
In October, we attended AoIR 2025 and participated in a panel on AoIR Ethics in Action: Guidelines, Regulations, and Practices Across Diverse Global Communities. We shared our knowledge and perspectives on ethical and privacy-aware research with online communities rooted in a recent whitepaper on the topic.
In October, we attended WikiConNA 2025 (Wiki Conference North America) and presented on the topics of Moderator Motivations and the Tone Check tool. We further hosted a session on NPOV from a research lens and supported the community in reflecting on how to improve NPOV practices rooted in scientific research findings.
In October and November, we gave two presentations about WMF’s AI Strategy for Editors: one as part of WikiConference Seoul 2025 and the other as part of itWikiCon 2025 (Italian Wikimedia Conference), the community conference for the Italian-language Wikimedia projects.
In November, we attended the Wikimedia research conference in Ukraine 2025. We introduced WMF’s Research team and its work to the Ukrainian research communities and raised awareness about resources available to the Ukrainian Wikimedia researchers. (Video)
Office Hour Series
We held over 25 one-on-one office hours to support researchers by answering questions about proposed or ongoing projects, guiding dataset access and analysis, sharing insights on Wikimedia Research team initiatives, exploring potential collaborations, and more. (Learn more)
Research Showcases
Our Monthly Research Showcases featured diverse presentations on Wikidata, Readers, and the impact of LLMs on knowledge production communities. (Learn more)
Wiki Workshop
We began organizing the 13th edition of Wiki Workshop and published the call for contributions. (Learn more)
Research Fund
We launched the fifth annual Wikimedia Research Fund cycle and invited established researchers to propose research projects or community building events. (Learn more)
Mentorship through Outreachy
We continue to participate in Outreachy and are excited to mentor a project focused on developing a tool that automatically generates prioritized micro-tasks for Wikipedia articles to help organizers and new editors. (Learn more)
Fellows Program
We relaunched the Wikimedia Research Fellows program, a year-long affiliation for post-PhD, early career researchers who are ready to make Wikimedia projects a sustained focus of their scholarship. (Learn more)
Collaborations
The Research team's work has been made possible through the contributions of our past and present formal collaborators. During the last six months, we established the following new collaborations:
- Akhil Arora is an Assistant Professor of Computer Science at Aarhus University where he leads the CLAN research group. Akhil is joined by Ankita Maity, a first year Computer Science PhD student. Ankita’s research interests are on responsible AI. Akhil and Ankita are collaborating with us on developing a systematic understanding of the usage of maintenance templates on Wikipedia and developing models for article improvement suggestions. (Learn more)
- Kaylea Champion is an Assistant Professor of Computing & Software Systems in University of Washington Bothell. She is collaborating with us on developing a meta-method for analyzing the state of Neutral Point of View policy in Wikipedia across multiple languages. (Learn more)
- Aaron Shaw is an Associate Professor of Communication Studies and Sociology (by courtesy) at Northwestern University. He is a long time Wikimedia researcher with expertise in online communities, human-machine interaction, organizational behavior and data science. Aaron is working with us in support of the NPOV working group with a particular focus on developing a meta-method for analyzing the state of NPOV policy in Wikipedia. (Learn more)
To inquire about becoming a formal collaborator, please tell us more about yourself and your interests.
Events
-
Research Showcases
Every 3rd Wednesday of the month (virtual)Join us for Wikimedia-related research presentations and discussions. The showcases are great entry points into the world of Wikimedia research and for connecting with other Wikimedia researchers. Learn more -
Office Hours
Throughout the month (virtual)You can book a 1:1 consultation session with a member of the Research team to seek advice on your data or research related questions. All are welcome! Book a session -
Wiki Workshop
March 25-26, 2026 (virtual)Join us for the 13th edition of Wiki Workshop, the annual forum bringing together researchers and practitioners exploring all aspects of Wikimedia projects. Learn more -
Research Track at Wikimania
July 21-25, 2026 (Paris, France + virtual)To mark Wikipedia’s 25th anniversary, we are collaborating with members of the Research Community to co-organize an engaging Research Track at Wikimania 2026. More information will be shared in the coming months.
For other Wikimedia community organized events, check the events calendar.
Trends to watch
Experimentation
Controlled experiments have helped the Wikimedia Movement establish causality and make research informed decisions. For example, research has demonstrated that if we help people see how their knowledge matters for Wikipedia, they are more likely to step in and help grow the projects (WMF-RAY 2025 winner: paper); or that we can increase Wikipedia’s pageviews by improving the Wikipedia article content on English Wikipedia (WMF-RAY 2021 winner: paper). However, the number of research controlled experiments on Wikipedia remains low relative to the Movement’s needs. In our conversations with researchers, we have learned that further incentivizing researchers for this type of research can help. The incentivizing can happen through a variety of initiatives: providing an experimentation infrastructure to researchers, reducing the barriers for conducting Wikimedia community aligned and ethical research on the projects, allocating funds to this particular type of research, and sharing some of the most important research questions Wikipedia faces today for researchers to learn about and consider working on. We will continue monitoring and working on this space including receiving input from the Wikimedia communities, hosting sessions with experts in the field (Video), and more.
Trust
As we engaged with the Wikimedia communities over the past months, trust and how to determine trustworthiness in a world in which AI generated content and AI operators have significantly increased remained an important topic. The Wikipedia model is based heavily on building trust between humans. While Wikipedia has a substantial number of bots operating on it, for the past 25 years those bots have operated primarily through processes in which the editors and humans behind the bots were known. LLMs introduce significant pressures to this model. As a research community, we face the question of what the future trust model for Wikipedia can look like? We encourage you to familiarize yourselves with the ongoing conversations in the communities (example, and another example) and reach out to us if you want to explore more.
NPOV
Neutral Point of View is a core and non-negotiable Wikipedia policy. When it comes to contentious topics, Wikipedia editors and users with extended rights are under significant pressure to uphold the implementation of NPOV on different Wikipedia languages. However, as we share in a whitepaper, Wikipedia communities have access to only a limited number of rigorous research on the topic and at the same time, they frequently run into old and new questions in their work. How can we, as the Wikimedia research community, support Wikipedia editors and users with extended rights in their work upholding NPOV on Wikipedia?
If you have ideas for running controlled experiments to improve Wikipedia, we invite you to a conversation in one of our office hours or to submit a proposal to Wikimedia Research Fund.
Donors
Funds for the Research team are provided by donors who give to the Wikimedia Foundation, and by a grant from the Sloan Foundation. Thank you!
Keep in touch with us
The Wikimedia Foundation's Research team is part of a global network of researchers who advance our understanding of the Wikimedia projects. We invite everyone who wants to stay in touch with this network to join the public wiki-research-l mailing list.
In addition, our team offers one-on-one conversation times. You are welcome to schedule one of our upcoming Public Office Hours.
We look forward to connecting with you.