Jump to content

WikiCredCon 2026/Notes

Add links
From Meta, a Wikimedia project coordination wiki

Use this page for pointers to important tools, pages, projects, people and organizations during the conference.

Resources

[edit]

Notes from the Resource Building Workshop

[edit]

Here are the stickies and notes created around the Resource Building workshop.

This was originally created on a virtual sticky board, and so there may be a few transcription issues. Also, these notes have not been completely reviewed for correctness, and names and other identifying information have been removed when not part of an already publicly available resource.

These notes are meant to capture what was discussed in April 2026 during WikiCredCon. If you see a note that has been added that is yours and would like to add some corrections/clarifications, feel free; please respect other people's notes though by leaving them there or by not overwhelming the tables. Thanks!

How do we address today's issues around manipulated content and reliable sources?

[edit]
What are the issues you are seeing or thinking about?***
Category Issue Examples Side notes
Source-level Source is unreliable (doesn't meet WP:RS standards)
Sponsored/PAID SITES as "Legitimate" Source
  • Pay to play publishers / self-publishing to achieve "notability-washing"
  • sources largely 'owned' by entities that serve private interests
Hacked AI/websites as path to potential Wiki/data poisoning https://www.bbc.com/future/article/20260218-i-hacked-chatgpt-and-googles-ai-and-it-only-took-20-minutes Related to above examples but not for payment, just because one can: "I can eat more hot dogs than any tech journalist on Earth"
News sources not disclosing use of AI -- should these be considered less reliable now?
  • Related: journalists & news orgs are not educated about reliable sources criteria --> awareness would incentivize them to be transparent about AI use and how to remain reliable
Zombie source: Repurposed previously legitimate news sites eg Ashland Daily Tidings: https://www.opb.org/article/2024/12/09/artificial-intelligence-local-news-oregon-ashland/ Also related to issue above. Also: Wayback machine could track ownership information as meta data that allows bots to push that information back to Wikipedia either directly to article or as talk page comment on legitimate sources.
False verification through source Source does not verify the information or topic (whether or not people use AI to create articles, but problem is scalar now) OKA book example: https://www.404media.co/ai-translations-are-adding-hallucinations-to-wikipedia-articles
Degradation of formerly reliable source e.g. manipulation of scientific information

https://www.pbs.org/newshour/health/cdc-vaccine-safety-webpage-changed-to-contradict-scientific-conclusion-that-vaccines-dont-cause-autism

Disappearance of formerly reliable source e.g. government data
False verification through source Source does not verify the information or topic (whether or not people use AI to create articles, but problem is scalar now) OKA book example: https://www.404media.co/ai-translations-are-adding-hallucinations-to-wikipedia-articles
Anglocentrism of the scholarly record notation added: 1+
Source fabrication OR hallucination Source doesn't exist (whether or not people use gen AI, problem is scalar now) * https://www.nature.com/articles/s41598-023-41032-5
Feedback loop issue: AI citogenesis (AI citing AI citing AI...) https://en.wikipedia.org/wiki/Wikipedia:List_of_citogenesis_incidents when AI trains on wikipedia content and later wikipedia references that
literally grokipedia
Journalist education: Educate journalists to know about article histories and talk pages teach them about Wikipedia features that enhance transparency (traceability, history, talk page, etc), notation added: 1+
Lack of Reliable Breaking Sources: Increasingly few reliable Breaking News sources and local reporting, community-driven journalism
Threatened Academic Research: shrinking access to academic freedom, tenured positions, and intellectual agency shapes sources can be created and cited
Larger Q: source trust bedrock how can we objectively verify the facts in a source without a prior belief in another source?
Article-level Political profiles are so polished: a sign of influence on Wiki?
Bias - More content without sourcing - potential bias introduction (ref OKA)
Bias - recency, especially political topics After specific event, "when perspectives differ and one may have more latency or revisions than another" ...
Quality - Translation eg https://en.wikipedia.org/wiki/Wikipedia:Administrators%27_noticeboard/Archive378?ref=404media.co#h-OKA:_problematic_paid_translation_and_lead_rewrites_via_LLMs_across_thousands_of-20260119145400 annotation added: ➕
WMF level interventions on content is there a way that the relationship on certain content, the community, and the Foundation affects reliability on Wikipedia when it comes to legal actions? note legal contexts: e.g. EU GDPR, data subjects are recognized to have the human rights of rectification (correction) and erasure ('right to be forgotten'): https://eur-lex.europa.eu/eli/reg/2016/679/oj/eng
Verifiability in terms of citations and provenance
Cross cultural discrepancies in how readers interpret articles annotation: 1+
Cross-language issues: Contradictions between articles in different languages
  • discrepancies between same article in difference languages (different perspectives/translation errors), annotation: ➕
  • flag when introducing?
Author/Editor-level Coordinated State actions
  • Iran: Issue: Large-Scale Upload of Islamic Republic of Iran State Media to Wikimedia CommonsStarting January 2026, investigative reporting by NPOV Media (neutralpov.com), led by journalist Ashley Rindsberg: https://www.thefp.com/p/wikipedia-editors-are-helping-iran-rewrite-history, documented what appears to be a coordinated campaign to upload content from US-sanctioned Iranian state media outlets to Wikimedia Commons. What I observed: over 10,000 images and videos sourced from Iranian state media were uploaded to Commons, primarily through a single account. The content bore watermarks from official state outlets, including Tasnim News Agency; a US-sanctioned, IRGC-affiliated organization — and Khamenei.ir, the official website of Supreme Leader Ali Khamenei.The material included Khamenei-branded propaganda videos, staged street interviews, professionally produced protest footage, and content featuring chants of "Death to America" and threats directed at the US President. One collection featured imagery with red HUD targeting markers and the statement "We will not let them go," drawn from a Khamenei speech, effectively a documented threat of retaliation against Iranian protesters.
  • An editor states that the a specific language Wikipedia is heavily edited by the government
  • annotation added (for general issue): ➕
  • NOT JUST BIAS: What I think is worth sitting with is this: a specific language-speaking community has largely stopped trusting Wikipedia as a neutral source. That trust, once lost, is very hard to rebuild. People aren't just frustrated with individual edits, they assume the whole editorial environment is compromised. That's the real urgency here. Every day this goes unaddressed, the platform loses credibility with exactly the audience it should be serving.
Editor-training: Opacity in editing decisions - difficult for newer volunteers to understand how to make articles reliable or what the issues might be "Processes for edits and revisions seem adhoc/opaque especially to newer editors: edits are regularly reverted by a very senior (25 years on the platform) editor, and there is seemingly no way offered to update the information -- it's a process violation, rather than an attempt to add an enhance knowledge creation.....verifiability on the platform often seems ad hoc and atomized, and the processes and mechanisms for ensuring or reviewing claims and citations is often opaque for casual -- but regular -- volunteer editors like my friend, who is more concerned with enhancing the encyclopedia than following the process."
  • Mentorship is important
Editor training - sourcing: Editors modifying text without a new/updated source or source does not verify new text.
Bias in editorship - how do we ensure we also attract new types of editors?
  • authors being biased by their background
Author-reader relationship - POV of the authors different from the readers
Conflicts of Interest https://en.wikipedia.org/wiki/Wikipedia:Wikipedia_Signpost/2026-03-31/Disinformation_report https://en.wikipedia.org/wiki/Wikipedia:Wikipedia_Signpost/2026-02-17/In_the_media
Non-human editors: Detecting whether user account is a bot https://www.404media.co/an-ai-agent-was-banned-from-creating-wikipedia-articles-then-wrote-angry-blogs-about-being-banned
Editor privacy and self-censorship
Systemic-level Wikimedia's Overall Credibility with more and more credible news sources blocking AI bots (an less credible sources allowing them), credibility in Wikimedia becomes even more important as it becomes a bigger share of what AI ingests.* https://arxiv.org/abs/2510.10315* https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/sr_25-07-22_ai_summaries_2/
  • The cultural narrative that everyone can edit Wikipedia = Wikipedia is not reliable needs to change. How do we shift the mainstream perspective to show that even though Wikipedia is freely editable, it still has rigorous checks and standards?
Defending Wiki Credibility How can we combat the narratives that try to draw an equivalence between Grokipedia and Wikipedia, i.e. the "I have my sources, you have yours" both-sidesing.
Quality inconsistency: within and between articles, non-uniform knowledge
  • also across languages -- Reliability knows no language barrier
  • Freshness v/s quality of info on Wikipedia. How to balance?
Content gaps: if there are major content gaps, it undermines credibility as users land on wikipedia pages
  • Different language Wikipedias have different content gaps (areas, not just articles); is this something we need to be concerned about?
  • Do Wikipedia's information and knowledge gaps widen amid rising authoritarianism and lack of government transparency?
Rise of (bad) LLM text as negative, not anti-tech
  • I'm really concerned about the rise of llm-generated text and what that means for the credibility of the sources we cite. Hallucinated information is already making its way into scientific papers and courtrooms. I'm also concerned about there not being enough pushback against hype in a fact-checking sense ("CEO says a thing journalism"). People opposed to AI are not opposed to technology or "in the way of progress".
Inequitable access to resources Access to frontier LLM tools (that produced good text) creating divide b/t community groups
Trust versus Identity
  • Multiple credible authorities agreeing on the same fact makes it credible
Technical issues Inconsistencies between technical makeup of different sources

***note: are these issues interrelated? yes. feel free to add connections among them if you see them or want.


How do or can we address the issues?
Category Issue Examples Side notes
Solutions (current and proposed) Core Policies and Practices - AI
  • Core policy change regarding the use of AI in content production, 20 Mar 2026
Core policy - updates https://en.wikipedia.org/wiki/Wikipedia:WikiProject_Policies_and_Guidelines
  • How do we do that?
  • Who is working on this?
  • how do we reflect changes in the policy downstream
Ecosystem education: Ensure that public interest media and news orgs understand how their work is interdependent with Wikipedia
  • educate journalists about Wikipedia core principles and the features that enhance transparency (traceability, history, talk page, etc) 🔥
Improved Tech -- AI/Claude kill switch https://en.wikipedia.org/wiki/Wikipedia:Writing_articles_with_large_language_models
Improved Tech -- Edit check https://www.mediawiki.org/wiki/Edit_check
  • "Put policies at its point of use"
Improved Tech -- AI models
Improved Tech -- Claim related https://phabricator.wikimedia.org/T399642
  • Technical solutions for pulling citations from sources
  • Tracking the provenance of a claim has become more important than ever
  • Automated detection tools to generate flagged citation lists -> Use AI to detect attestations not supported by citations.
Governance Innovation Employ decision-making tools like pol.is that surface more nuanced consensus more efficiently and civilly (better vibes!) * annotation added: ❤️
Citation Database
  • quality signals on sources https://arxiv.org/abs/2603.17146,
  • Cybersecurity People Ideas: Merkle TREES 🔥, NLP analysis of Keyword alignment vs historical (from archive.org), Automated review if business ownership changes or legal filings made (review public disclosures automatically), Flag citations of websites whose domain owner has changed
  • Makes it easier for editors (annotation added: 1+)
  • connections between sources, languages, fields, reliability checks, verifiability, flagging of claims
  • WikiCite! annotation: ❤️ (2+)
  • Who is going to maintain a centralized citation catalogue?
Community Efforts/Practical Guides https://en.wikipedia.org/wiki/Wikipedia:WikiProject_AI_Cleanup Guides to detect AI writing: https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing, FREEDOM OF EXPRESSION, ALGORITHMS, AND COLLABORATIVE KNOWLEDGE THE WIKIMEDIA MODEL AS AN ALTERNATIVE.pdf (guide from Wiki Argentina): tiny.cc/mup1101 Page at https://meta.wikimedia.org/wiki/Anti-Disinformation_Repository needs to be updated to help people learn what other people are working on
Existing Tools for Reliability Bias analyzer: https://nethahussain.github.io/wikipedia-bias-analyzer/, Cite Unseen - roll it out to all editors
  • Improving tooling, making scalable systems for the Wikimedia community
Collecting lists
  • Collection of reliable and unreliable source lists: https://meta.wikimedia.org/wiki/Cite_Unseen/sources ||
  • Let's make a combined cross-language list of 'reliable sources' or of 'unreliable' ones. For now such lists are in different languages and formats. Annotation: ✅
Bias - attracting new editors Source doesn't exist (whether or not people use gen AI, problem is scalar now)
  • potential solutions: - make it more obvious that there is an in-person community you can join - meet people where they are, e.g. mobile editing
  • Understand better what motivates people and adapt their experience to that
  • Community norming: how do you balance being welcoming with policy introduction
Perennial source discussion expansion?
  • We need a structured, community-edited "questionable sources list" that has a lower threshold than the deprecated sources list, spam blacklist, or perennial sources list
Journalist education: Educate journalists to know about article histories and talk pages teach them about Wikipedia features that enhance transparency (traceability, history, talk page, etc), notation added: 1+
Volunteer reference desk?
  • If there are only a few who have time/inclination, more potential for burnout.
Library level protocols?

Additional comments:

  • Not unrelated to things like this, noted: https://firstdraftnews.org/articles/fake-news-complicated/
  • Prompt to think about whether motivations were helpful to identify when it came to solutions didn't go anywhere. Some that were initially offered:
    • Discrediting (falsely)
    • Wreaking havoc
    • Increased stats
    • Commercial promotion
Retrieved from "https://meta.wikimedia.org/w/index.php?title=WikiCredCon_2026/Notes&oldid=30515440"