Please collaborate on this draft essay of ideas for the future of Wikipedia - hoping to turn into an essay or two depending on feedback

While research does not yet seem to be a strength of AI tools, perhaps a tool could be custom-made for Wikipedians that focuses on reliable sources? Should not be too difficult to build a tool for editors that is more comprehensive and extensive than Wikipedia:Advanced Source Searching#Custom search engines which have a WP:Overreliance upon Google and may no longer work after January 1, 2027 when searching >50 websites.

Wiki survey algorithms/implementations could help us surface consensus in large discussions with lots of participants (such as elections and other bigger decisions made by the Wikipedia community). Anonymized surveys can help editors to speak their mind more freely and openly (avoid groupthink), encourage more participation by providing a healthier discussion environment and help discover where among all the back-and-forth there is actually sufficient consensus to act upon. This also has a side-benefit of reducing the risk of being labeled as ‘social media’ which is increasingly coming with age-gating and other legal and policy challenges. Lastly, this is foundational to attempts to address bias of various kinds in Wikipedia, including by expanding the pie of editors (articles that have been subject to more scrutiny by more perspectives tend to have less bias).

Additionally, the Wikimedia Foundation exerts enormous control over the project and seems like a risky concentration of power that has led to numerous conflicts. I worry about the willingness of Wikimedians to speak truth to the kind of power that awards grants along with controlling a lot of other levers of the various projects. I worry that this grant-making power leads to a natural soft capture of affiliated organizations, especially in an environment where the Foundation has refused to allow new affiliates to apply for recognition for the last couple years. I think that the grantmaking aspects and the endowment should be split off from the Foundation and exist as a separate entity, one whose board is mostly elected by editors and where the Foundation has to apply for grants (above say a flat 30-50% of fundraising) just like any other affiliate. This separation of powers also helps to diminish the conflicts of interest that are growing with tech companies, especially from them paying for API services from the Foundation.

1. Fear that the AI skepticism among many editors is a form of unilateral disarmament amidst a global information war, where adversaries of verifiable/neutral knowledge are able to use the latest technologies but its defenders are not.
2. Worried that there will be too much nostalgia/love for the hand-crafted process of Wikipedia by humans so that it becomes a place for knowledge artisans who enjoy hand-crafting articles, long past the time when we are a convenient source of reliable information.
3. Given that only 0.6% of Wikipedia articles are considered Good articles, how do we get that number up to 3% let alone 6% of articles without substantial use of AI at some point? What are our goals for the % of articles that are rated good articles? (ostensibly many articles do not have enough coverage to qualify as a good article?)
4. If AI is so expensive that Wikipedia would have to limit its use, why not prioritize it for content where editor safety is most at risk: controversial topics, especially where transnational repression, stochastic terrorism, and other risks are known?

Apparently AI is good at short summaries.^[1] Some short-term applications in Wikipedia could include:

• Edit summaries
  • EdiSum^[2]
• Talk page discussions^[3]
• Lead paragraph
• Maybe even an article summary

Reported to be a valuable use of AI from Wiki Education students.^[4]

Wiki Education recommends AI for learning how to access sources as well as for suggestions for which sources to search in.^[4]

Wiki Education^[4] and Jimmy Wales^[5] recommend this use case. This includes suggesting categories or identifying other missing features of an article.^[4]

AI can help to suggest articles that would be a good match for an editor based on their skills and interests.^[4]

• …where there is low press freedom or academic freedom on the topic in much of the world
  • Also, documenting trends on freedom of speech and academic freedom could help better understand which topics would be most valuable to prioritize.
• where sites run by and for powerful people, like Grokipedia, are least likely to have a useful entry

As more and more people get information from less transparent sources (social media, AI chatbots, etc.), Wikipedia has a huge competitive advantage in being perhaps the most transparent source of information on the internet. By focusing on creating and improving content that is disputed and distributed a lot in low-transparency environments, it could bring back appreciative users to the site and maybe even help to grow the editor base.

Articles on sources are especially helpful due to how contentious they can be as well as how helpful they are for Wikipedians deciding how much weight to give that particular source within Wikipedia on any given topic.

Certain WikiProjects like Freedom of speech and Human rights likely have a higher proportion of articles on contentious/political topics.

This idea is a credit to a presenter at Wikimania 2025 who has started to focus on articles that do not exist in other languages.

If tools like WikiChat, an AI bot that refines its answers by drawing from the Top 25 Wikipedia languages, become more available this strategy makes more and more sense to focus less on translation and more on quality in whichever language is most likely to support the highest-quality article.

AI tools, if they are going to be helpful in specific tasks, benefit from high-quality training data and examples. Bringing articles up to these standards might make lots of tedious tasks a thing of the past more quickly than they otherwise would.

The other benefit is it gives editors time to come across issues around guidelines and interpretations of rules that can help to clarify and improve the standards that Wikipedia will use in the future.

Curious what the conversations have been between MDWiki and Wikipedia:WikiProject Medicine but until this fork has resolved, it seems less impactful believing that it will have to be merged at some point (especially given the decline in editors across either project). If MDWiki wants added restrictions for editors to medical content to merge the fork, then that may be a worthwhile compromise.

Other questions about MDWiki.org vs. Wikipedia’s medical coverage:

1. is there any academic or independent analysis of MDWiki? Wikipedia has so much research done on its health information that it seems (as of September 2025) like the best place to contribute.
2. there has never been a successful fork of Wikipedia - what would make MDWiki any different?
3. can the contributors/members of MDWiki (990), especially James Heilman (User:Doc James) and User:Ozzie10aaaa be reconciled with Wikipedia?

Security is increasingly going to be an important differentiator for retaining and attracting users and editors as the internet becomes increasingly unsafe. The Foundation should do a complete redesign of its backend software as soon as a more secure infrastructure becomes apparent, for example, and probably should double its focus on security.

If visiting websites becomes too unsafe for editors at some point, would it make sense to work towards building or accessing a search index that only includes text from reliable websites that can be used to build Wikipedia?