Wikipedia's AI Policy: LLMs Banned for Article Creation, Exceptions for Copyediting and Translation

Wikipedia's AI information page codifies community consensus on AI use. The core rule is straightforward: LLMs are prohibited for generating or rewriting article content. Two exceptions exist: basic copyediting and LLM-assisted translation (per WP:LLMT), which requires the editor to be fluent in both source and target languages and to manually review every output before publishing.

Key Policies and Their Shortcuts

• G15 (WP:G15): Speedy deletion for pages “that could only plausibly have been generated by large language models without reasonable human review.”
• WP:AIIMAGES: Bans images wholly generated by AI, with specific rules for biographies of living persons (WP:AIIMGBLP).
• MOS:AIUPSCALE: Prohibits AI upscaling software on images.
• WP:RSML: AI-generated content cannot be cited as reliable sources. LLMs are listed as perennial sources.
• WP:AITALK: Talk page comments obviously written by an LLM may be struck or collapsed.
• WP:NICETRY: Unblock requests that appear AI-written are likely summarily rejected.

Practical Implications

If you're building AI agents that interact with Wikipedia, this means:

• Do not submit LLM-generated articles without thorough human rewriting — they risk immediate deletion under G15.
• Translation tools (like those using LLMs) are allowed only if the editor fully controls both languages and reviews each sentence.
• AI-assisted image generation is off-limits for Wikipedia articles.

The policy page also notes that works created solely by machines are not copyrightable (as of 2025), which is a relevant legal consideration for Wikimedia Commons uploads.

Discussion Timeline

The page tracks the evolution of these rules: initial discussions began in December 2022 in response to ChatGPT, with a disclosure suggestion in March 2023. The policies have been updated through March 2026.