Repository navigation
feat: support blocking AI training crawlers - #349
Conversation
🤖 MERGED
GitHub merged this pull request.
571d2373-1733-4630-a54c-4eedce6cb4dd Selected findings run after merge and open separate pull requests. |
commit: |
|
| Module | Client gzip | Server raw |
|---|---|---|
| @nuxtjs/robots | 608 B | 173 kB |
Package files (6)
Rows overlap and include unused code. These totals do not measure deployed output or npm downloads.
Excludes dependency files, types, source maps, and other non-code files.
| Package files | Summed gzip | Raw | Δ gzip |
|---|---|---|---|
| @nuxtjs/robots · export . | 7.1 kB | 27 kB | 🔴 +59 B (+0.8%) |
| @nuxtjs/robots · export ./content | 292 B | 503 B | — |
| @nuxtjs/robots · export ./util | 466 B | 784 B | 🟢 -6.2 kB (-93.0%) |
| @nuxtjs/robots · built code files | 43 kB | 141 kB | 🔴 +1.6 kB (+3.9%) |
| @nuxtjs/robots · app source files | 3.3 kB | 7.2 kB | — |
| @nuxtjs/robots · server source files | 9.8 kB | 24 kB | — |
Dependencies (8)
Declared dependencies. Their installed size does not show runtime cost.
| Package | Dependency | Requested |
|---|---|---|
| @nuxtjs/robots | @fingerprintjs/botd | catalog: |
| @nuxtjs/robots | @nuxt/kit | catalog: |
| @nuxtjs/robots | @vueuse/core | catalog: |
| @nuxtjs/robots | consola | catalog: |
| @nuxtjs/robots | defu | catalog: |
| @nuxtjs/robots | nuxt-site-config | catalog: |
| @nuxtjs/robots | nuxtseo-shared | catalog: |
| @nuxtjs/robots | ufo | catalog: |
How this is measured
- Compare the same app with and without the module.
- Client: all emitted JS/CSS, summed per-file gzip. This includes lazy chunks.
- Server: deployed files, including external dependencies. Source maps are excluded.
- Default module settings, Node server preset. Memory and request speed are outside this report.
Base: main @ 3bf98b6 · 2026-10-12 · differences below 16 B are ignored
This release includes breaking changes. Run a manual major release. Automatic releases support patch and minor only. |
🔗 Linked issue
Resolves #347
❓ Type of change
📚 Description
Site owners cannot opt out of AI training through
blockAiBotswithout also blocking AI search and answer agents.Adds
blockAiBots: 'training'for a curated list of training crawlers and training opt-out tokens. Dedicated search and answer agents stay unblocked by this option. Existing robots.txt and indexing rules still apply. Thetruesetting keeps blocking the complete existing list.The module owns the list locally, with operator documentation beside each token. The docs explain that Google-Extended also controls Gemini grounding, and CCBot supplies a shared archive.