Wikipedia editors have adopted a new way to remove low-quality content that is written by large language models. The change expands the existing speedy deletion process so admins can act without a week-long discussion when pages show obvious signs of AI generation. This step is meant to stop fake citations and boilerplate LLM phrasing from staying on the site.

What the new rule allows

Under the updated criteria, administrators can delete new articles and major rewrites quickly if the text shows clear LLM telltales. Examples include phrasing that reads like a canned output or citations to papers or sources that do not exist. The change is an extension of the speedy deletion policy that already handles clear nonsense and spam. The project that helped create the wording is known as WikiProject AI Cleanup.

Why volunteers moved to this faster process

Editors and admins reported a surge in quickly generated submissions that contained errors and fake references. Many of those pages used formulaic language that showed no sign of human review. The volume strained the normal week-long review process. Volunteers said a faster tool was needed to stop the worst examples before they spread. The move follows other community pushback against planned AI features that editors said would lower quality.

How administrators judge an article

Admins look for a pattern of warning signs. The triggers include citations to nonexistent journals or books and stock phrasing such as a preamble that reads like a direct LLM prompt. When those signals are present and the page is new or has a big rewrite, the admin can apply the speedy deletion criteria. Deletion is reversible if someone shows that the content is valid. Editors stress that the rule targets the clearest bad cases and not content that a human simply wrote poorly.

Limits and concerns

Some volunteers call the new rule a stopgap. They agree it helps with the most obvious junk. They also worry that quick deletion could remove useful work that only needs cleanup. The Wikimedia Foundation does not set policy for individual projects, so community debate shaped the rule. Past experiments with AI summaries faced strong resistance and were shelved. The new tool sits alongside other efforts such as Edit Check and the WikiProject AI Cleanup guides.

What this means for readers and contributors

Readers should see fewer pages that cite things that do not exist. Contributors who use AI to draft text still need to check facts and add real citations. Editors encourage users to cite original sources and to mark content that used AI tools. If a page is deleted in error, it can be restored after review. The deletion policy and the Signpost notes explain how restoration and appeal work.

How to spot AI slop on Wikipedia

Look for oddly phrased intros and references that do not match the topic. Check the citations. If a source looks unfamiliar, search for it outside Wikipedia. If you find clear errors, report them on the article talk page or flag the page for review. Volunteer editors will check and clean up articles that are salvageable. When a page is beyond repair, admins can remove it quickly under the new criteria.

Why this fight matters

Wikipedia is widely used as a first stop for research. Low-quality AI-generated pages can spread misinformation and create false leads. The volunteer community has long relied on open editing as its main defense. The new rule is a tool that helps the community act faster when large language models create obvious junk. It does not replace careful review, but it does reduce the time the worst content remains visible.

Wikipedia editors have given themselves a faster way to clear out the most obvious AI-generated slop. The change is narrow by design, and it aims to protect the site while volunteers work on longer-term fixes. The community will keep refining the rules as new problems appear and as tools evolve.