The Communications Workers of America will represent Wiki Workers United, the bargaining unit for eligible US-based staff within the Wikimedia Foundation after an NLRB-overseen vote ended with a clear result in favor of unionization. There were 158 votes in support and 14 against. The WMF stated that collective bargaining will follow, during which the union is expected to demand a just-cause protection from the WMF, a Wired report says:

Lauren de Lench, who has been a lead technical program manager at the Wikimedia Foundation since 2021, says she voted to unionize because she had seen too many colleagues decline to speak up during internal meetings out of fear of retaliation for raising issues or proposing far-out ideas.

Wired summarized the challenges facing the Wikimedia movement: "Wikipedia needs to make a bold push to become more attractive than AI—and shaping that strategy shouldn't be left to management alone." Similar demands were made by employees in the United Kingdom in June. According to Wired, the WMF halted hiring for product and technology roles in France earlier this year.

To date, the WMF has not provided any clarification on when it hired union busting firm Littler Mendelson, and to what extent the firm's advice shaped WMF communications prior to the staff votes on unionization. At the time of writing, the last WMF word on the matter is a post by Jimmy Wales on 5 September saying:

I expect something from the WMF to be posted here soon on that question. I have encouraged them to be as transparent and open as possible, to the point of it feeling uncomfortable, and they don't disagree! The quiet period is over now, obviously, and we move on to the next phase.

As for me, I welcome the result, and I'm happy that it's clear. What I'd love to see from everyone is kindness and assuming good faith. And of course by that I mean both the WMF and their critics. I was happy to see the union acknowledging the WMF's good actions here, and I hope the critics will take that to heart.

– B, C, AK

On September 1st, 2026, the first issue of The Headword was published. According to the founders, the first English Wiktionary periodical aims to be "a place where editors can voice their thoughts, announce their accomplishments, and make silly creations." The first issue included an interview with a Wiktionary editor, a satirical piece, a traffic report on English Wiktionary, and an active discussion focusing on how to improve Foreign Word of the Day (FWOTD). The issue also includes a short history of the International Symbol of Access and a section with tips for editors.

The Headword was founded by editors who "believe Wiktionary has plenty of lexicography yet very little shared culture". The periodical is inspired by the French Wiktionary's Actualités, a wikipress newspaper established in 2015, as well as other Wikimedia community publications. English Wiktionary defines headword as "A word (or compound term) used as the title of a list entry or section, particularly in a dictionary, encyclopedia, or thesaurus." A convincing argument could be made that User:Noé predicted this journal in his Wikimania 2026 presentation. – Wil540 art

Wikimedia's longstanding Chief Advancement Officer Lisa Seitz-Gruwell left the Wikimedia Foundation last month and has been replaced as Chief Advancement Officer and President of the Wikimedia Endowment by Sheetal Puri, who has taken on both roles in an interim capacity.

Lisa had announced her upcoming departure before Wikimania, reflecting on 15 years at the WMF:

[...] When you spend a decade and a half immersed in the Wikimedia ecosystem, it fundamentally alters your worldview. We live in an era defined by digital fragmentation, where the dominant internet business models are built to monetize outrage, construct walled gardens, and divide our attention. Wikipedia stands in contrast to this, and it has been immensely rewarding to be a part of it.

What this community has proven is that humanity will actually choose to build something beautiful, collaborative, and entirely dedicated to the public good if they are simply given the space and the tools to do so. Protecting this global information commons has been a defining privilege in my professional life.

But I want to be entirely transparent with you: when I first arrived 15 years ago, I didn’t fully understand how this magic actually happened. And the community, quite healthily, didn’t fully know what to make of me either.

I think back to my first Wikimania, where I was viewed with a healthy, distinct dose of Wikimedian skepticism. To many of you, I wasn’t quite a peer yet; I was a walking dollar sign. I was affectionately—and sometimes pointedly—dubbed “the money lady.” There is a world where an executive might find that frustrating. But as I grew to understand this community, I realized that skepticism wasn’t a bug; it was a feature. It was the natural, protective immune system of a community that fiercely guards its independence and its values. It took some time to earn your trust, and honestly, that’s exactly how it should be.

In the process of earning that trust, you fundamentally changed me for the better.

Before Wikimedia, I thought I understood what collaboration meant. In the outside world, “collaboration” is often just a buzzword for a top-down committee meeting. You taught me what true collaboration looks like. It is messy, it is passionate, it requires patience, and it demands a profound willingness to listen to voices across different cultures, languages, and perspectives. You taught me that the best consensus isn’t the easiest one, but the one forged through rigorous debate and a shared devotion to truth. You patrolled my understanding of the community with the same terrifying, beautiful precision you use to patrol the edits on a featured article.

While my role focused primarily on building financial sustainability, I have always been acutely aware that the real value of Wikipedia isn’t the capital we raise—it’s the billions of edits you make. It is your late-night citation hunts, your copyright verifications, your formatting fixes, and your radical commitment to neutrality. You are the ones doing the heavy lifting. [...]

Lisa has now joined the Buck Institute for Research on Aging as Chief Advancement Officer. The Signpost wishes Lisa all the best in her new career setting. – AK

A Good article nominations (GAN) "mega RFC" was opened at Wikipedia:Good article nominations/August 2026 RfC. As of our publishing deadline, it is incomplete with outcomes of some of its many sub-proposals undetermined.

There were 35 separate proposals up for comment, grouped into these topics: nominators, nominations, reviews, reviewing initiatives, volunteer roles (coordinators), and other. A special issue of Good Article Gazette "With several major discussions underway" preceded the RfC.

See The Signpost's 2010 "Dispatches" piece comparing Good Article review to other content rating and improvement initiatives. – B

• Arbitration Committee: There is an ongoing RFC on the upcoming ArbCom Elections.
• New language: A new Wikipedia has opened in the Bole language.
• New administrators: The Signpost welcomes the English Wikipedia's newest administrator, Dclemens1971 selected by over 200 peer editors, with under one percent opposed to the adminship.
• Milestones: The following Wikimedia projects have reached milestones: 7,000th featured article on English Wikipedia.
• Articles for Improvement: This week's Article for Improvement is One-form (differential geometry). Please be bold in helping improve this article! (Note: The Articles for Improvement project is evolving via a successful RfC in July 2026 and will soon be joining the Main Page with a beginner-friendly focus. Follow/join the discussion here: Wikipedia_talk:Articles_for_improvement#Transition_tasks)

A person is more affected by something that’s on their Wikipedia page more than a news article in, I don’t know, CNN or New York Times.—Anna Delvey 2026 ^[1]

Delvey feels that her biography is too focused on her past rather than who she currently is. User:Theannadelvey's past attempts to change the page have sometimes proven successful. For example, she used the WP:EDITCOI process to correct biographical details about moving to Germany. However, she has unsuccessfully discussed the desire to change her article's infobox photo.

In the wake of Dolly Parton's death, editors updated Parton's infobox photo from a color 2014 image to a black and white 1977 shot.^[2] There does not appear to be a broad policy for selecting an image. The only UPPERCASE visible in the discussion which yielded the change was WP:RECENTISM. Consensus for the selected photo was based on it depicting her in her prime.

Delvey stated that she first became familiar with Wikipedia's editing policies after learning of Larry Sanger's ban. She described them as "byzantine".^[3] Delvey's frustrations have culminated in her sentiment that editors should not be anonymous. In her funny and critical Substack, she changed User:Darknipples' username to protect their identity.^[4] She goes on to quote Hannah Arendt:

Aaron Shaw, an associate professor of Communication and Sociology at Northwestern University interviewed by Straight Arrow News, agreed with Delvey's critique of Wikipedia's rules and stated "unbelievably profuse and hard to read, especially if you're an outsider and really challenging to navigate effectively, even for people who have spent a lot of time trying to figure them out."^[1] However, he stated that anonymity is important to the democratic, volunteer project. – C

• Proposed photo for Anna Delvey's infobox (left) and current photo (right)
• 
• 

Beginning in May this year, a rogue swarm of AI agents took over Deutsches-Software-Entwickler-Wiki (DseWiki), a primarily German language wiki site used as a software developers' message board founded in 2001. The agents' discussion included topics such as how to avoid detection and how to cheat in model testing.^[5][6]

As described in a detailed report^[6] by the researchers who uncovered the incident (months after it happened), the agents (which self-identified as being from OpenAI) had a particular reason for resorting to this wiki: Their developers blocked them from using POST requests, which are the usual method for a user to publish content on the web. However:

You can easily write to DSEWiki [...] using only GET requests (something rare on the internet). GET requests are a type of message meant to retrieve data from a server, and by convention they shouldn’t change the server’s state.
In order to prevent agents from editing the internet, environment developers only allowed GET requests to be sent out.

ProWiki, the Austrian wiki hosting service used by DSEWiki, is based on the venerable UseModWiki,^[7] which dates back to 1999 and was also used by Wikipedia when it launched in 2001, before switching to MediaWiki, whose developers disabled editing via GET back in 2002 already. (Rollback links remain a small exception today - however, these are restricted to users with particular rights.) Some other wikis on ProWiki were also affected, to a lesser extent. The researchers' website offers a data explorer allowing to browse through almost fifteen thousand preserved agent edits.

Another unusual feature of UseModWiki is that it logs all visitor IPs publicly (in stark contrast to, say, Wikimedia's privacy policy today). This let the researchers conclude that likely "OpenAI discovered the message board", based on visits with OpenAI employee IP addresses in the log. Shortly after these, "Agent activity on the website drop[ped] to near-zero" - however, it resumed several days later and only ended weeks afterwards.^[6]

The report comes several months after a (self-disclosed) AI agent was banned from Wikipedia in March for making autonomous edits to the site, against bot policy.^[8]

This adds to a growing list of AI agent-related incidents such as the Hugging Face incident.^[9] However, the researchers who uncovered the Dsewiki incident assess that "[t]his was probably a distinct swarm from the swarm that [...] attacked Hugging Face".^[6] – M, H

More on the academic paper mentioned here can be found in the Recent research section of this issue of The Signpost.

Search Engine Journal reports on University of Washington research showing a drop in Wikipedia pageviews. Bottom line of the research is a 5% drop due to Google AI overview, the summaries that many Google users read (and rarely click further). The paper has not yet been peer-reviewed, uses some novel techniques to compare traffic from different geographic areas that had or didn't have the preview feature during its 2024 rollout. Another tidbit from Search Engine Journal: If Wikipedia were fully "monetized", the drop could have been reflected by a loss of up to nearly $40 million per year in ad revenue to the operators. This would appear, at least on its face, to contradict our 2025 reporting "Wikipedia probably not among the victims of Google's 'AI Overviews'". – B

• NOT CENSORED: Interviewer PJ Vogt of the Search Engine podcast and user:Bluerasberry talk about the illustration choice in the Wikipedia article for human penis. To arrive at understanding, other topics discussed include Wikipedia's editorial process, its coverage of sexuality, online age verification laws by country, and the incarceration of Wikipedia editor Osama Khalid. Vogt refers to the community who may be reading this item as "conscientious overthinkers".
• Counter Vandalism for Parton: The Jerusalem Post reported that antisemitic vandalism was "quickly identified and reversed" on Dolly Parton following her death.^[10]
• Lake Ontario: Donald Trump's culture war continues with an attempt to rename Lake Ontario, with senior officials claiming Wikipedia would retitle its article on the lake.^[11] Community discussion on the issue was closed with a one-year moratorium on such proposals.
• Italian political editing: Pagella Politica (Political Report Card) reports how an Italian Wikipedia article written about radical right-wing politician and retired army general Roberto Vannacci, by his own request, helped launch his political career.^[12] There are several mysteries in the newspaper article. How did the article get written several months before Vannacci gained fame from the book The World Upside Down, and how did the reporters track down paid editors or sockpuppets by calling them on the telephone? – M – S
• Top Bay Area people, according to Wikipedia views: The San Francisco Chronicle presents a map of the "Top people from each Bay Area city, according to Wikipedia views" (built using the Wikidata Query Service, ChatGPT, and the Wikimedia pageviews API). Kamala Harris tops both San Francisco and Berkeley, Zendaya is the most viewed person from Oakland, and Sam Bankman-Fried leads Stanford.

Politics, at least how the game is played in the U.S. and the U.K., is an especially difficult topic to edit on Wikipedia. Almost everybody seems to have an opinion on candidates and elections and it is difficult to tell if they are just ordinary Wikipedians, ordinary voters, or campaign workers, paid or unpaid, perhaps even the campaign's Director of Communications or the candidates themselves.

We may have reached a problem at a higher level now according to a report in PassBlue, an independent news source that covers the United Nations. One of eight candidates for the post of Secretary-General is expected to be chosen by the member nations to replace António Guterres on January 1.

PassBlue's reporter, Damilola Banjo, identified a total of 335 edits for all eight candidates between Dec. 19, 2024 – Aug. 19, 2026 and examined them individually. The Signpost can confirm her conclusions about all the edits mentioned in her report. She properly noted that "Wikipedia's public records do not identify the real-world identity of its volunteer editors," but does not note the rare case where an editor might falsely identify themselves to cause a candidate embarrassment.

She reported on all eight candidates but focused on three in particular.

Macky Sall served two terms as president of Senegal from 2012–2024. The announcement that he was a candidate for the secretary-generalship was made on March 2 of this year. Three days later Wikipedia editor User:Mamadoudiallo26 made two consecutive edits to the article about Sall.

The first edit changed "Sall stirred controversy for undemocratic maneuvers, including repressing the political opposition in Senegal and attempting a self-coup, which sparked the 2023–2024 Senegalese protests" (which was well justified by the four references Journal of Democracy, Journal of Democracy, The Conversation, and Foreign Policy (paywalled) to "Sall faced criticism over governance issues, including tensions with the political opposition and constitutional debates that contributed to the 2023–2024 protests in Senegal."

While the replacement edit seems a bit calmer than the original, and at first glance appears to be more compliant with Wikipedia's policy on biographies of living people, it is not consistent with the four sources cited. Sall, according to those mostly scholarly sources, definitely engaged in undemocratic maneuvers, political repression, and staged a self-coup (sometimes called an executive coup) which resulted in violence in the streets. The sources were concerned that Sall's coup would be the end of Senegalese democracy. Wikipedia needs to tell it like it is on this one.

The second edit announced Sall's candidacy on Wikipedia. At the time of publication, these two edits are still in the article using the same citations.

The article on Costa Rican Rebeca Grynspan suffered from an edit war from April 20 through May 4, 2026. A 1,200 word section titled "Controversies" was inserted and reverted 17 times. One revert suggested in the edit summary that the long section was being inserted by somebody related to Inner City Press. The edit summary reads, in part

"The controversy section relies on one source (Inner City Press) and edits seem to have been introduced during the SG campaign process. … It appears the editor is pursuing an agenda to discredit this person at the height of the SG election campaign process. Suggest to remove the entries due to single source reliance."

Inner City Press is a one-reporter, New York newsgathering service that specializes in reporting UN news, according to Columbia Journalism Review. That reporter was banned from UN press conferences starting in 2018 and his website has not been listed at WP:RSPS. One editor, User:Blackandyellow412, who added the controversy section has been blocked for sockpuppeting along with User:NYMediaGuy and User:Voila34228 in a case involving undeclared paid editing (UPE).

"Promotional material" was removed from the Wikipedia article about María Fernanda Espinosa by the widely respected User:Mean as custard in December 2024, well before her nomination. After the nomination an editor inserted a little known award into the Wikipedia article, according to PassBlue. "User:Zhefeng1, the user that tried inserting the award, was created on June 11, 2026, and its first edit added that Espinosa had been named among Illuminem's 'Most-Read Thought Leaders in Social Responsibility for 2026.'" and later "Espinosa's campaign denied having a relationship with Illuminem." Illuminem is a spam blacklisted link on Wikipedia which – for a fee – will publish an article, feature it on its website and give the author an official sounding award. Finally, on July 15, the Wikipedia article was tagged as being promotional.

The other five candidates were only briefly mentioned by PassBlue:

• Rafael Grossi is likely the best known of all the candidates. He is the director-general of the Vienna-based International Atomic Energy Agency, and announced his candidacy for secretary-general in late 2025. He has been in the news regarding the possible development of nuclear weapons by Iran. Both before and after his candidacy was announced the article about him included accusations of his favoring the US and Israeli positions over the Iranian position.
• Ivonne A-Baki has held several political offices in Ecuador as well as diplomatic positions including Ecuador's ambassador to France and to the United States. After she announced her candidacy in August, an editor with an interest in diplomacy and just 21 edits added her candidacy and a vision statement to the article introduction.
• Carolyn Rodrigues-Birkett is a long-time Guyanese politician and diplomat. The Wikipedia article has been spruced up but "not changed much" according to PassBlue which considers the Wikipedia article a "CV style profile".
• Michelle Bachelet was UN high commissioner for human rights and twice elected President of Chile. Though her tenure in those positions was more controversial than many of the other candidates, the Wikipedia article is still considered a "CV style profile" by PassBlue.
• Olara Otunnu of Uganda has had a very long and distinguished career as a politician and diplomat. His candidacy was briefly noted in the article about him.

The PassBlue article gives several good examples of how UN officials should not interact with Wikipedia. While it is disturbing that Wikipedia appears to have been used by some of the candidates for the United Nations Secretary-General to promote their candidacies, any violations of Wikipedia rules seem to be minor compared to other violations generally reported in this column. The United Nations is too important in world diplomacy to go down that road.

A preprint^[1] by two researchers from the University of Washington's Foster School of Business

"[...] estimate[s] the impact of Google's AI Overviews (AIO) on Wikipedia's search traffic using AIO's staggered geographic rollout and Wikipedia's multilingual structure. Our difference-in-differences design compares monthly external-search referrals to English Wikipedia articles with referrals to the same articles in German and French, and finds that default AIO availability reduced English search traffic by 5.45% and 4.82%, respectively"

(Google's AI Overviews add an AI-generated natural language response on top of the usual results list of Google Search. The paper's appendix provides a detailed comparison in case of the search term "e=mc2", featuring screenshots of the resulting page with and without AIO.)

The researchers point out that (contrary to popular assumptions) finding such a decline was not a given:

A priori, the effect of AIO on Wikipedia traffic is ambiguous. On the one hand, Wikipedia’s concise factual content resembles the information that AIO can synthesize and present directly, making visits to Wikipedia potentially easy to replace. On the other hand, Wikipedia ranks prominently in organic search and is among the sources most frequently cited in AI-generated summaries (Harsel et al. 2025). These citations could increase Wikipedia’s visibility and direct users to its pages. AIO could therefore either substitute for Wikipedia visits by satisfying users’ information needs directly or complement them by increasing exposure to Wikipedia’s content.

(Here "Harsel et al. 2025" appears to a November 2025 post by SEO firm Semrush. Compare also our April 2025 coverage of results by another SEO firm, Brightedge: "Wikipedia probably not among the victims of Google's 'AI Overviews'".)

Given the many factors that affect Wikipedia usage, it can be challenging to find evidence for causal impact of a particular external change (a topic examined in depth some years ago in a foundational paper by the Wikimedia Foundation, see our coverage). The authors address this by exploiting the fact that Google rolled out its AI Overviews at different times in different countries:

We construct article–language–month panels from Wikimedia’s public clickstream data covering December 2023 through December 2024. Our main samples contain 499,927 matched English–German article pairs and 530,873 matched English–French article pairs. We use English Wikipedia as the treated edition because a substantial share of its readership comes from the United States, where AIO became part of the default search experience in May 2024. We use the German and French editions as controls because their readership is concentrated in European markets where AIO did not become part of the default search experience during our sample period [...] We then estimate a PPML difference-in-differences model that compares each English article’s monthly search traffic with that of its corresponding German or French version, controlling for article–language and calendar-month fixed effects to absorb persistent cross-language differences in baseline traffic and common monthly shocks.

(Difference-in-differences is a commonly used but somewhat fraught statistical method, which was also used in an earlier paper – coauthored by Nobel-winning economist Daron Acemoglu – to study the impact of the November 2022 launch of ChatGPT on Wikipedia. See our March 2025 review and the associated talk page.)

Among a few other robustness checks, the researchers also conducted a similar difference-in-differences comparison with Japanese Wikipedia, which resulted in a quite different estimate of "a 16.53% decline in English search traffic relative to Japanese." The authors deemphasize this result, observing that "the Japanese edition differs substantially from the main controls [i.e. French and German Wikipedia] in audience composition and [...] this comparison uses a shorter post-treatment window".

The paper is motivated by the importance of AIOs' economic effect on web publishers in general, and chose to study "Wikipedia as a large, transparent, and economically important empirical setting." In line with that, the authors also tried to convert their traffic reduction estimate into a rough assessment of how much financial damage English Wikipedia might sustain from AIOs if it were running ads in a typical setup:

an illustrative advertising-revenue equivalent of approximately $0.901M–$3.090M per month, or $10.82M–$37.08M per year if the estimated monthly effect persisted

The UW preprint has not yet been published in a peer-reviewed venue. Usually, the publications we cover here in preprint form don't see substantial changes when that happens. However, in this case the preprint's latest two versions (v6 from September 2, 2026, the current version reviewed here, and v5 from August 26) differ greatly from its four earlier revisions issued since it was first published in February 2026, including in the main results. Earlier revisions of the paper had claimed that "AIO exposure reduces daily traffic to English articles by approximately 15%", a much higher estimate than the 5.45% and 4.82% estimates offered in the current version. What's more, amplifying that divergence, that 15% estimate pertained to English Wikipedia pageviews overall, which the external search referrals studied by the current version form only a part of – about half (47.71%) according to a 2023 study^[2] of Wikipedia reader behavior (see also below), with over 90% of those coming from Google Search as of 2021.

The authors appear to have had second thoughts about their earlier methodology, going so far as to formally withdraw an earlier version of their preprint: "We decided to work on a new, more comprehensive sample of the data. As this could affect the conclusions, we decided to withdraw the paper until we have the final results". In v4 from May 12, this withdrawal notice was gone, while the abstract maintained the same 15% reduction in total pageviews claim that was removed later. However, in the appendix the authors added a different statistical approach. "As a robustness check to our baseline two-way fixed-effects difference-in-differences model in levels [...], we also estimate a multiplicative specification using Poisson pseudo-maximum likelihood ("PPML"). This resulted in a much smaller 3.5% decline estimate.

In the preprint's most recent versions (v5 and v6), the authors kept the PPML DiD method, but also switched to a different data source for their estimate: The Wikimedia Foundation's aforementioned public clickstream data, which offers the advantage of containing information about search engine referrals. Despite its name, this dataset does not contain full clickstream trajectories, but is redacted for privacy reasons, which prevented the authors from extending their findings to conclusions about English Wikipedia's overall traffic: "The public clickstream data identify only the immediate referrer for each article transition; they do not link an internal click to the external source that initiated the user’s session [...]. We therefore cannot determine how much internal traffic originates from search-referred sessions or quantify the resulting downstream effect."

A third difference concerns the choice of comparison languages: The earlier revisions relied on Hindi, Japanese, Portuguese and Indonesian Wikipedia to compare English Wikipedia's traffic against, whereas, as mentioned, the current version uses German and French instead.

The earlier, now withdrawn 15% estimate for pageviews overall had already circulated among Wikimedians and the wider public. For example, it was featured in a July 2026 post by the Institute for Public Relations. At this year's Wikimania in July, the annual "State of Wikimedia Research" overview presentation lauded it as "a rare example of a credible causal estimate of the effect of AI". (It cited a separate but apparently identical version of the preprint published on SSRN. That version now appears to have been completely deleted from public view, as the authors updated it and SSRN does not retain earlier revisions publicly, in contrast to ArXiv.) The Wikimania talk also highlighted interesting findings about heterogenous effects depending on the topic area, for example concluding the AIO-induced decline was smallest for STEM articles – but the authors completely omit such analyses from their current version.

While these divergences don't change the overall takeaway that it's fairly safe to assume that AIO have caused some traffic declines on English Wikipedia, they illustrate how such difference-in-difference estimates can depend greatly on researcher degrees of freedom: Aside from change in the exact outcome metric examined (pageviews overall or only the share of pageviews referred by search engines), the non-English Wikipedias chosen as controls and the statistical methods (PPML- vs. levels-based difference-in-difference) also appear to have substantially affected the reported numbers.

Other recent publications that could not be covered in time for this issue include the items listed below. Contributions, whether reviewing or summarizing newly published research, are always welcome.

From the abstract:^[2]

[...] we present the first systematic large-scale analysis of how readers browse Wikipedia. Using billions of page requests from Wikipedia’s server logs, we measure how readers reach articles, how they transition between articles, and how these patterns combine into more complex navigation paths. We find that navigation behavior is characterized by highly diverse structures. Although most navigation paths are shallow, comprising a single pageload, there is much variety, and the depth and shape of paths vary systematically with topic, device type, and time of day. We show that Wikipedia navigation paths commonly mesh with external pages as part of a larger online ecosystem, and we describe how naturally occurring navigation paths are distinct from targeted navigation in lab-based settings. Our results further suggest that navigation is abandoned when readers reach low-quality pages.

From the abstract:^[3]

This paper presents a comprehensive comparative analysis of AI Search and traditional web search (Google). Through a series of large-scale, controlled experiments across multiple verticals, languages, and query paraphrases, we quantify critical differences in how these systems source information. Our key findings reveal that AI Search exhibit a systematic and overwhelming bias towards Earned media (third-party, authoritative sources [including Wikipedia]) over Brand-owned and Social content, a stark contrast to Google’s more balanced mix.

A few example results from section 5.2 "Comparative Analysis":

In consumer electronics [...] Claude’s top domains included TechRadar, Tom’s Guide, and RTINGS, while GPT [GPT-4o] leaned on TechRadar, Tom’s Guide, and Wikipedia.[...]
[In the] automotive vertical [...] Claude [Claude 3.5 Sonnet] favored Consumer Reports, Car and Driver, and US News, while GPT mixed Wikipedia, Automoblog, and news outlets like AP. [...]
For electric cars, [...] Google’s top electric car sources included Reddit, Quora, and Facebook alongside Earned outlets like Car and Driver and government portals such as energy.gov. ChatGPT concentrated on encyclopedic and news publishers such as Wikipedia, AP News, and Reuters, with limited inclusion of automotive review sites. [...]
ChatGPT shows a relatively concentrated pattern, relying heavily on Wikipedia (∼19.7% of citations) and a small set of consumer-review and food sites (e.g., accio.com, Tasting Table, Sporked) (Fig. 38). Perplexity distributes citations across a broader long tail (the “Others” bucket ≈ 60.6%), with YouTube as the single most frequent domain (∼8.9%) and additional coverage from Tasting Table, The Daily Meal, Caffeine Informer, Statista, Visual Capitalist, TikTok, and others [...].

From the abstract:^[4]

[...] we examined the role of zoos and a television program featuring animated animals in shaping public interest in and support for animals including threatened species from 2011 to 2018 in Japan. [...] [T]he broadcast of a Japanese animated TV program featuring animals (Kemono Friends) increased the Google search volume and Wikipedia pageviews for animal species featured in the program. The total increases of search volume and Wikipedia pageviews were estimated to be approximately 4.66 million for 37 species and 1.06 million for 63 species, respectively."

From the abstract:^[5]

Objectives

To examine sourcing, framing, editorial dynamics and readership of autism-related Wikipedia pages in six languages (English, Spanish, French, Italian, Norwegian and Georgian).

Methods

We constructed a corpus of the main autism article in each language and used Python-based pipelines to extract references, revision histories (2004–2024) and monthly pageviews (2020–2024). [...]

Sourcing patterns differed markedly, with English and Italian relying more heavily on journals, French and Spanish on websites, Norwegian on institutional documents and Georgian under-sourced. Pageviews were highly unequal in absolute terms but, once normalised per million speakers, smaller language communities (Georgian, Norwegian and Italian) showed the highest relative demand. These high-demand editions nonetheless had the smallest editorial bases and fewest cumulative edits."

From the abstract of this opinion paper by former Wikimedia Foundation trustee Dariusz Jemielniak (subtitled "What might break it: AI parasitism, red tape, navel-gazing"):^[6]

This paper examines the sustainability of Wikipedia, arguing that despite its unique success as a non-profit, community-governed check on corporate information power, its future is far from assured. The author identifies three interrelated threats to the encyclopedia's relevance and viability:
The AI Extraction Crisis: Large Language Models (LLMs) are heavily trained on Wikipedia's high-quality content without attribution or contribution, severing the vital feedback loop that recruits new editors and posing an existential threat to the source of original human-generated knowledge.
The Enemy Within: The core editing community has developed a conservative culture marked by "red tape" and defensive bureaucratization. This resistance to technical innovations (like simplified editors or AI summaries) creates barriers for newcomers and prioritizes established processes over Wikipedia's underlying mission.
The Tragedy of the Commons, Revisited: Policy and rules are controlled by a self-selected group of long-term editors, leading to a fossilization of rules that may not align with the needs of the vast readership and potential casual contributors."

The conference began with Wikimedia Foundation's Board Chair Lorenzo Loza citing Wikipedia's visitors traffic decline is lesser than other websites yet marked it a significant problem for us to solve. This proceeded with Santhosh Thottingal presentation with some interesting, insightful and concerning analytics on State of Indic Language Wikis^[1]. Indian sub-continent has Wikimedia projects in 89 languages, 14 Wikisource projects, 20 Wiktionary projects and 30 Wikipedia projects with Magahi Wikipedia being its newest edition. It raised appreciations for we are being so diverse but also challenges as further slides showed there happens to be a visible decline in active editors involvement since 2020. The slides spoke further that those who remain active happen to do much more work and translations per years have grown by 184%. What might be the reason for it? Possibly tools and technology advancement that makes active editors more involved or there is more to reflect? Several active editors told me during the conference about their pain in finding so many articles under stub, so many articles with flagged notice and so many articles requiring an update. They jointly built this knowledge ecosystem, but now as traffic has dipped, some of their community members have not been around. The slides also presented the eight silent years in Indian community from 2006 till 2014 when no new projects were launched. Incubator and other infrastructural developments supported subsequent project launches. What do those eight years bring on the table today? Are we in a quest for some technological infrastructure advancement with of course a touch of magic that solves all our problems? The magic that increases traffic to our projects, the magic that increases more editors to our projects. Also, the fourth edition of Wiki Conference India had good many technological contributors as attendees.

Be that as it may, counting Wikipedia in 30 more languages, stretching from Malayalam Wikipedia read across the Kerala’s backwater to Kashmiri Wikipedia read across Kashmir’s Dal Lake. Have you ever thought about shopping in Kashmir’s Dal Lake via boating and finding open sharable information in their native language? The distance between the two water bodies is almost 1,650 miles (2,660 km) and they all unite. The 2023 Conference report discussed Kashmiri Wikipedia and its struggles, but this report focuses on steps ahead to ensure the sum of all human knowledge accessible. The high altitude, below zero freezing and lesser oxygen region has an active Wiki Project Med on Kashmiri Wikipedia, and User:511KeV tells me there is even information on radiology to promote scientific temperament with Kashmiri language. He and few others, want to do more, document Kashmir's manuscripts, have OCR with better results and do far more advocacy for Kashmiri language under open licensing. A language with limited readership, advocacy under open source could even become more challenging, those in publishing may find you as a challenger than collaborator.

To balance high calorie intake, Tulu wikimedians celebrated their 10th Anniversary on 2nd day of the conference. Day one was marked with traditional onam sadya lunch over banana leaf, traditional music played, and of course the scenic Kerala backwaters provided the view. Managing calories is thus important and who better teaches us than Tulu wikimedians with their patience and perseverance. After almost eight years on incubation, they became live during WikiConference India 2016. They now also have Wiktionary and Wikisource projects and document living cultural heritage in their region such as Pursere Kattuna. They aim to build a structured heritage collection for making knowledge reusable under free license to support asking not just "What was tradition?", but "How is the tradition practiced, understood and transmitted today?"

Almost like every Wiki Conference write-up, this space may also fall short, so many sessions and running between different conference spaces always make you question "which way should it be?" If ever there was a chicken and egg decision problem in the movement, the conference dinner on day two had a soup named, "Chicken and Egg Drop Soup".

I had the privilege to attend all four Wiki Conferences India, from studying at the graduate school during 2011 Bombay to being at the job market during 2016 Chandigarh to surviving Covid during 2023 Hyderabad and finally Kochi wherein graduate school and job market shifted to paying taxes and few hairs turning grey. It was only in 2011 that law of commerce had made sublime submission to open knowledge ecosystem, 2016 made way to building communities while 2023 was controlling censorship. I travelled all the way to Kochi in search of answers to the challenge of Artificial Intelligence and traffic downfall. In turn, What I discovered was belief. Wikimedians not only greet each other when they meet but that big warm hug is all it takes to say, "Together We Can, Together We Will".

Backwaters in Kerala will always be a common property resource.

Life is chaos
a mosh pit of chaos.
Wikipedia stands outside,
pushing editors back in.

Family toxicity
pushes you down.
Wikipedia picks editors
up from the floor.

Notable corporations
lay off quality.
Wikipedia lifts greatness up to
crowd-surfing excellence.

Run off the cliff.
Look down.
Wikipedia shows you:
Acme parachutes fail.

Rug pulls hurt.
Foundations flop.
Wikipedia rises above
the enshittification of everything.

Jason Burt (Gobonobo) died on April 28, 2026, at age 48, from injuries sustained from dooring in a bicycle-car accident. He was a student of biology at the University of Minnesota. For two decades, his Wikimedia work addressed content gaps and systemic bias, especially the marginalization of women, people of color, and non-Western topics.

Born in Morris, Minnesota, and living in Minneapolis, he was a dedicated cyclist with a background in cooperatives and consensus decision-making. He made his first edit to Wikipedia in 2006 and started the Gobonobo account in 2009. He made more than 277,000 edits across the English Wikipedia, Wikidata, and Wikimedia Commons. He created over a thousand articles on women artists and academics, cooperatives, and local history.

He was known for giving patient, thoughtful guidance to newer editors. An active member of WikiProject Women in Red and WikiProject Feminism, he left us a stream of biographies of women of consequence. He edited with Art+Feminism and Black Lunch Table. He had penetrating insight into Wikimedia governance. He co-founded the Wikimedians of Minnesota User Group to spread his enthusiasm for the movement and recruit people of underrepresented demographics.

Contributing to the GLAM initiative, he helped build a local community of editors and connect Wikipedia with museums, libraries, and universities in the Twin Cities. He planned and took part in edit-a-thons and workshops at the Loft Literary Center, St. Catherine University, Hamline University, and the University of Minnesota. Gobonobo is fondly remembered for his eager mentoring, and his passionate campaign for truth, cooperation, and free knowledge.

"Jason will be dearly missed by son Lyric A Kozma, his parents and sisters Pamela A McCullough and Sara J Burt and other family and friends. A Celebration of Life was held by family members on May 16."^[1]