Product & Tech Progress on the Annual Plan

News from the WMF

Product & Tech Progress on the Annual Plan

Ana Eira is a Movement Communications Specialist at the Wikimedia Foundation. Part of the content of this post was previously published on Diff.

Every year, the Wikimedia Foundation, together with Wikimedians across the projects, creates an Annual Plan that outlines our priorities for the year ahead. Since July, we've been working under a new plan that puts a strong focus on Product and Technology. In this post, we'll share some highlights of the progress we've made in the first two quarters of the fiscal year 2024–2025.

Wiki Experiences

In Q1, we released the highly anticipated Dark mode, which had been a longstanding wish shared by many Wikimedians. This feature reduces eye strain for readers by providing a low-contrast reading experience. We have now deployed Dark Mode to over 40 wikis, including for anonymous users.

Another longstanding request was a way for contributors to easily find on-wiki events and communities that interest them. In Q2, this resulted in the expansion of the Collaboration List feature in the Campaign Events extension, which now includes a tab for "Communities". This tab features WikiProjects on the local wiki, as well as links to the WikiProject page and the associated Wikidata item. The Collaboration List was also recently enhanced with more search filters, so people can now search for events by wikis or topics. Additionally, wikis with the CampaignEvents extension enabled can access the Invitation List feature, which was released in Q1. This feature helps organizers identify users who might be interested in their upcoming event or WikiProject, based on their contribution history.

Automoderator is the latest deployment in a suite of tools that volunteer editors can use to fight disinformation while making content moderation more efficient. A substantial number of edits are made to Wikimedia projects which should unambiguously be undone, reverting a page back to its previous state. In the past, patrollers and administrators had to spend a lot of time manually reviewing and reverting these edits. Powered by the new "Revert Risk" machine learning models, Automoderator automatically detects and reverts harmful edits. Between Q1 and Q2, Automoderator was deployed on six new wikis.

In addition to that, AbuseFilter editors and maintainers can now make a CAPTCHA show if a filter matches an edit. This allows communities to quickly respond to spamming by automated bots. In the same spirit of supporting editors on the wikis, the Chart extension, which enables editors to create data visualizations, is now available on MediaWiki.org and three pilot wikis.

We have started introducing updates aimed at enhancing the editing experience on the iOS mobile app as part of a broader transition from a reader-focused platform to one that fully supports both reading and editing. Editors on the iOS mobile app have now updated navigation features, which include a new Profile menu that allows for easy access to editor features like Notifications and Watchlist from the Article view. Expanding on this effort, the Alternative Text suggested edits feature has now been fully deployed to production on the iOS App for Spanish, Portuguese, Chinese, and French Wikipedias. This feature is designed to help newcomers add alt text to images, aiming to improve accessibility and engagement.

We are also continuing to look for ways to improve the experience for contributors with extended rights on the projects, like stewards. Stewards can now specify if global blocks should prevent account creation. Before this change, all global blocks would prevent account creation. This will allow stewards to reduce the unintended side-effects of global blocks on IP addresses.

To guarantee a more secure and privacy-conscious experience without compromising convenience, we are updating Single User Login, the system that allows users to login on one Wikimedia site and be automatically logged in across all, to ensure compatibility with browser anti-tracking measures and will roll out to all users by the end of March. This will also further improve account security by limiting all authentication to a single domain.

The Foundation launched a series of community conversations with Wikimedia Commons volunteers and stakeholders to help prioritize support efforts for the next fiscal year. While current efforts focus on improvements to the UploadWizard that will make it easier for moderators, there are a wide variety of interests and needs across Commons that have been discussed with and across community members in the last few months.

We've resolved a total of 650 volunteer-reported issues in Phabricator in the last 6 months. For instance, on multilingual wikis, users can now hide translations from the WhatLinksHere special page. We have also resolved bugs in important areas like Add a Link, the Android Wikipedia App, and the "Download as PDF" system, among others.

To foster close, ongoing collaboration with community members across WikiExperiences work, the Foundation launched a Product and Technology Advisory Council (PTAC) as a one-year pilot in October 2024. The council brings together technical contributors, affiliates and the Foundation to collaborate on building a more resilient, future-ready technological platform. The Council is charged with making recommendations around product and tech development work over a multi-year horizon to help align to the strategic direction of the Foundation. PTAC has already released its first recommendation.

Future Audiences

We are responding to volunteer requests to experiment with how new generations of people may read and use Wikipedia content. Our reader experience teams are running six experiments now to help us learn where to invest even more time and attention. To learn more about "free knowledge everywhere" as a potential strategy for multigenerational sustainability, we began an experiment to remix Wikipedia content into short "fun fact" videos and publish them on TikTok, Instagram, and YouTube – popular platforms where younger generations like to spend time and learn. Over the course of October 2024 to January 2025, we attained a total of 1.87M views and 1.51M unique users reached via this experiment, indicating that this is an effective strategy to bring Wikipedia knowledge to younger audiences at scale. We are continuing to test different content types and are inviting community members who are active on these platforms to join in and collaborate with us.

These are just a few examples of the progress made in the first two quarters. To learn more, read the full update on Diff.

← Previous "News from the WMF"

In this issue

1 May 2025 (all comments)

Disinformation report

News from the WMF

+ Add a comment

Discuss this story

These comments are automatically transcluded from this article's talk page. To follow comments, add the page to your watchlist. If your comment has not appeared here, you can try purging the cache.

== "fun fact" videos (Future Audiences) ==

[...] we began an experiment to remix Wikipedia content into short "fun fact" videos and publish them on TikTok, Instagram, and YouTube – popular platforms where younger generations like to spend time and learn. Over the course of October 2024 to January 2025, we attained a total of 1.87M views and 1.51M unique users reached via this experiment, indicating that this is an effective strategy to bring Wikipedia knowledge to younger audiences at scale.

I'm glad WMF is working on bringing Wikipedia content to new and younger audiences. But I'm curious about the meaning of "effective" and "at scale" here.

For context, during those four months, Wikimedia projects attracted over 60 billion human pageviews (the vast majority of them for Wikipedia). So this WMF effort increased Wikipedia's readership by less than 0.005% during that time.
Now, perhaps these videos enabled us to reach people who had never before heard of Wikipedia and caused them to become regular Wikipedia readers (possibly for the rest of their lives). But do we have any evidence that this has happened "at scale"?
Lastly, m:Future Audiences/Generated Video doesn't provide any information about how much effort and donor money went into creating these videos, i.e. about efficiency - how much did we need to spend to generate those 1.87 milion views? I know that the Foundation is not in financial distress of course, it can (and to some extent should) afford experiments. But there has been a call (including from the current CEO IIRC) to focus its resources more on the most impactful work, and this report leaves one with doubts whether that includes this kind of video content production by WMF staff.

HaeB (talk) 23:33, 1 May 2025 (UTC)[reply]

Hey @HaeB, happy to provide more context on your two questions:

On the cost: The Future Audiences team (which at the time was one software developer and me) was able to create our homebrewed video generation tool and the first few months' worth of video content during a 1-week focused hackathon last August. We approached this project as "creating a self-serve tool for our External Comms team" precisely so that it would be as cost-effective as possible – i.e., require as little time/labor to create as many videos as possible – and External Comms has continued to use the tool to generate more videos on more topics. In addition to these videos, you'll also see some that were made by the agency our External Comms team works with on other social content (e.g., Wikimania content) – they've been helping us create different kinds of content (e.g. videos featuring community members) to see how this performs as compared to our little DYK explainers. Having worked on many WMF software projects over the years, I can confidently tell you this is some of the scrappiest software development I've personally been involved in (as are all FA projects, by design). (For scale: I think the comparably scrappy things at WMF were some of the Editor Engagement Experiments from Ye Olden Days of 2013-4, and even that was a bigger & higher-resourced team than FA!)
On scale: The "scale" in the snippet you quote refers to the size of the audience we were able to reach and the time to get to that number relative to kinds of social campaigns we've done over the years. Getting these kinds of numbers this quickly (let alone with the very minimal investment we put in) in the social space is, as far as I understand, unprecedented. But you're right, that while getting a few million views on social content is a big deal for us, it is absolutely not a large scale for these platforms! Successful creators can achieve many orders of magnitude more reach. We intentionally chose to start with the scrappy version of this work to learn quickly/cheaply, and based on the results we believe there is room for achieving true "at-scale" reach with additional investment into higher content quality/quantity next fiscal year.

HTH, and let me know if you have more questions! Maryana Pinchuk (WMF) (talk) 17:33, 2 May 2025 (UTC)[reply]

Thanks for the response and this additional information! It is indeed helpful, but it still leaves me a bit puzzled about several aspects:

during a 1-week focused hackathon / some of the scrappiest software development I've personally been involved in - I see, that makes sense, good to know that 1) this didn't take up an entire 7-person team for half a year and 2) appears to have been built as a tool with many future reuses in mind. That said, now I'm a bit confused why m:Future_Audiences#2024-25 seems to list this ("Generated Video") as one of only four projects for the entire year (from which a naive observer might have estimated that it took the team an entire quarter)? In any case, I guess we need to keep in mind as readers of these "Progress on the Annual Plan" posts that they are not a comprehensive activity reports, but rather just compile some highlights (as the intro transparently says).
Either way, I think it's great that WMF invests in experiments and that your team exists; one might even think that there should be more experiments of this kind. But any experiment should eventually be evaluated in terms of impact and costs of the thing being tested (including opportunity costs).
In addition to these videos, you'll also see some that were made by the agency our External Comms team works with on other social content (e.g., Wikimania content) – they've been helping us create different kinds of content (e.g. videos featuring community members) to see how this performs as compared to our little DYK explainers. - that sounds great, but why would one use the same metric to directly compare videos about our movement itself (such as Wikimania content) with those highlighting engaging information about arbitrary topics (even if attributed to Wikipedia)? That indicates a rather unreflected conception of impact, and a lack of a theory of change. As a concrete example, 10000 views for a video that helps bring in (say) 30 more Wikimania attendees from targeted audiences or 10 more high-quality program submissions for Wikimania are orders of magnitude more valuable for us than 10000 views for a video that presents information from a Wikipedia article about some non-Wikimedia conference. Or to put it differently: It will always be easier for Comms staff to find engaging "fun facts" among arbitrary Wikipedia topics (and often, well, more fun).
I am also kind of curious whether the team was aware of this past WMF research that we covered a while ago here in the Signpost/research newsletter: "An awareness campaign in India did not affect Wikipedia pageviews, but a new software feature did" (regarding a video that many Wikimedians found indeed very engaging and which gathered about 4 million views, but did not translate into measurable increases of readership on Wikipedia itself).
But you're right, that while getting a few million views on social content is a big deal for us, it is absolutely not a large scale for these platforms! - I guess that's true, but my comparison was actually to views on Wikipedia itself, not to views on those platform in general. And I still have to say that even if it can be maintained without further costs (and that would seem unlikely, given that the External Comms team presumably consists of decently paid staff), a persistent 0.005% increase in views of Wikipedia content (on external platforms and Wikipedia itself combined) would pale in comparison to say the estimated 1.4% increase we gained from this experiment a few years back (yes, I'm a bit biased there, but I do think WMF and others in the movement have long been too blase in assuming Wikipedia will always rank first on search engines and thus have neglected basic SEO too much). That would be a 280 times higher impact.

Regards, HaeB (talk) 04:27, 11 May 2025 (UTC)[reply]

What do you think of The Signpost? Share your feedback.

Home

About