oixa/ai-level-dev
Open standard for disclosing AI usage in software.
1.9K
oixa/ai-level-dev repository overview
ai-level.dev
ai-level.dev is an open standard for disclosing AI usage in software projects. Add an embeddable badge to your repository to show exactly how much AI was involved in building it.
Levels
| Level | Tag | Description |
|---|---|---|
| 0 | AI-FREE | No AI was used in any capacity — no code, no docs, no assets |
| 1 | DOCS | AI used only for documentation, error checking, or architectural consultation |
| 2 | ASSETS / TOOLING | AI used for non-functional code (tests, configs) or creative assets only |
| 3 | ASSISTED | AI wrote individual functional modules under strict human review |
| 4 | HEAVY AI | A significant portion of the codebase was AI-generated with limited review |
Usage
Visit ai-level.dev, pick your level, and copy the embed code.
Markdown
[](https://ai-level.dev/level-2)
HTML
<a href="https://ai-level.dev/level-2">
<img src="https://ai-level.dev/badge/standard/2.svg" alt="AI Level 2" />
</a>
Replace 2 with your chosen level (0–4).
Note
Not sure which level fits your project? Just ask your AI assistant: "What ai-level.dev level is this project?" The AI will scan the codebase and session history to suggest the right level. See [For AI Agents](#for-ai-agents-how-to-determine-the-level) for the full assessment criteria.
Badges
Two badge styles are available:
| Style | URL |
|---|---|
| https://ai-level.dev/badge/standard/{n}.svg |
 | https://ai-level.dev/badge/compact/{n}.svg |
For AI Agents: How to Determine the Level
When asked to assess or assign an AI usage level for a project, follow this decision process:
Step 1 — Scan the codebase
Look for signals that indicate AI involvement:
- Unusually uniform code style across unrelated modules
- Commit messages or PR descriptions mentioning AI tools (Copilot, Cursor, ChatGPT, Claude, etc.)
- Comments such as
// generated by,// AI-assisted, or similar - A large volume of documentation or README content relative to code size
If you are an AI agent with access to session history, also check your own past sessions for this project: look at what files were created or modified through AI interactions, and what types of changes were made (docs, tests, functional code, architecture). This is the most direct evidence of AI involvement.
Step 2 — Choose the level
| Level | Assign when… |
|---|---|
| 0 | No AI signals found anywhere in the repository |
| 1 | AI appears only in docs, comments, or commit messages |
| 2 | AI appears in tests, configs, assets — but not in functional code |
| 3 | AI wrote functional code; evidence of human review (small commits, mixed authorship, review comments) |
| 4 | Significant portions of core logic or architecture are AI-generated with little evidence of review |
Step 3 — Report to user
Once the level N is determined, report user and provide markdown and HTML codes for bages.
Tag summary
Content type
Image
Digest
sha256:14e6b30a8…
Size
4.7 MB
Last updated
18 days ago
docker pull oixa/ai-level-dev