How the ratings work

More and more open source projects publish rules about AI: some welcome it, some ask you to say when you used it, some ban it outright. This site gathers those rules in one place so you can check before you open that pull request.

How friendly is friendly?

Each project gets one rating for how friendly it is towards AI-assisted contributions. The rating comes from the project’s rules for code, with bans on AI-written issues or text pulling it down.

Friendly (AI-friendly)

Welcomes AI-assisted work. AI-assisted contributions are welcome under the same rules as any other contribution.

Mostly friendly (AI-friendly with conditions)

Welcome, with extra rules. AI-assisted contributions are accepted with extra rules, such as disclosing AI use, or AI is banned for issues or written text.

Unfriendly (Not AI-friendly)

Doesn’t accept AI-generated work. AI-generated contributions are not accepted; at most minor help such as autocomplete or grammar fixes.

No policy (No AI policy)

No policy yet. No published policy on AI-assisted contributions.

What we look at

Code contributions

Patches, pull requests and other changes written with AI tools.

Issues & bug reports

Bug reports, security reports and issues found or written with AI tools.

AI-written text

Pull request descriptions, issue comments and summaries written by AI.

Each area gets its own face on the same scale. Some come with conditions:

• DiscloseAI use must be disclosed
• AccountableContributor takes full responsibility
• Human reviewA human must review the output
• License-safeOutput must be license-compatible
• No agentsNo autonomous agents
• Minor help onlyOnly minor help (autocomplete, grammar)

Conditions that every good contribution meets anyway, like taking responsibility for your work, keep an areaFriendly. Extra hoops, like mandatory disclosure, make it Mostly friendly. Allowing only minor help makes it Unfriendly.

Where the data comes from

1. Someone suggests a project, ideally with links to its policy: a CONTRIBUTING file, a code of conduct, a docs page. An AI agent also searches for the project’s AI rules, and only uses official pages: the project’s own site, docs, repository or announcements, never news, blogs by others or social media.
2. We save a plain-text snapshot of each source in the repository, so every change is visible in git.
3. An AI model reads the full text and fills in the ratings. Each rating must come with a quote, and a script checks that every quote appears word for word in the snapshot.
4. A human reviews every change before it goes live.
5. Every week the sources are fetched again. If the text changed, the project is re-checked and reviewed.

Yes, we use AI to track AI policies. We think that’s fine, as long as a human signs off on every change.

Add or fix a project

Open an “Add a project” issue, with links to the project’s policy if you know them. Spotted a mistake? Use the “Suggest a correction” link on any project page.

For AI agents

Machine-readable versions of everything live at /llms.txt, /llms-full.txt and /projects.json. Every project page also has a Markdown twin at /projects/<slug>.md.

Related work

• open-source-ai-contribution-policies, a community list of AI policies.
• The Generative AI Policy Landscape in Open Source by Kate Holterhoff at RedMonk.

This is a summary, not legal advice. Projects change their rules and individual maintainers may be stricter than the written policy, so read the source before you contribute.