AI usage-preference declaration
What it checks
Where a site declares how AI systems may use its content. The IETF AIPREF work attaches that preference to a Content-Usage response header or a robots.txt rule and explicitly leaves the HTML head out of scope; the head-level noai/noimageai convention has real adoption but no documented consumer. The audit reports what a site declares and where, and never treats declaring nothing as a defect.
(The pre-rewrite description asserted that <meta name="ai-content-declaration"> is “how AI systems discover your llms.txt or AI usage policy” and that GPTBot and ClaudeBot read it. That was false; it is quoted with its refutation in the rewrite section below.)
Why it matters
A head-level AI declaration (<meta name="noai">, <meta name="tdm-reservation" content="1">, or an ai-generated declaration) causes AI crawlers to change training/ingestion behavior or causes AI systems to label the content. Falsifiable: no AI vendor recognizes any of these names, and the active standards work explicitly attaches preferences somewhere other than the HTML head.
Evidence
AI content declaration meta tags
Two distinct things live here and both are pre-consumer. (1) Opt-out declarations. The W3C TDMRep Community Group Final Report (2 Feb 2024) does formally define <meta name="tdm-reservation" content="1|0"> and <meta name="tdm-policy" content="URL">, positioned as a technical answer to EU DSM Article 4. But it is explicitly “not a W3C Standard”, and it names no implementing consumer. The DeviantArt-origin noai/noimageai convention has measurable adoption: 88,000+ domains as of June 2026, 87.8% of them via the meta-tag placement, meta adoption up 26.5% month-over-month (Originality.AI). (2) AI-generated declarations: there is no HTML head standard at all — IPTC’s Digital Source Type (trainedAlgorithmicMedia) targets the XMP packet embedded in image/video files or a C2PA manifest, not the page head.
Limits
Decisive. The IETF AIPREF attachment draft of 19 Aug 2026 is Standards Track, authored by Google and Mozilla, and updates RFC 9309. It defines exactly two attachment mechanisms: the Content-Usage HTTP response header, and a Content-Usage rule in robots.txt. It defines no HTML meta element and no link relation — ‘Embedded Preferences’ is acknowledged and left out of scope. So the standards trajectory is running away from the head, not toward it.
Google’s supported-meta-tags list omits noai and tdm-reservation; Originality.AI’s own study concludes “Major AI companies point elsewhere … rather than honoring the noai meta tag specifically.” OpenAI’s and Anthropic’s crawler docs document robots.txt only. Keep as experimental with a plainly worded caveat that these tags currently express intent with no known enforcing consumer, and track AIPREF for the header/robots.txt path — do not present them to users as protection.
Sources
- OpenAI crawlers and user agents — OpenAI, vendor-doc (verified 2026-08-21)
- Does Anthropic crawl data from the web, and how can site owners block the crawler? — Anthropic, vendor-doc (verified 2026-08-21)
- Associating AI Usage Preferences with Content in HTTP (draft-ietf-aipref-attach) — IETF AI Preferences (AIPREF) WG — G. Illyes (Google), M. Thomson (Mozilla), draft-spec (verified 2026-08-20)
- TDM Reservation Protocol (TDMRep) — Final Community Group Report — W3C Text and Data Mining Reservation Protocol Community Group (2 Feb 2024), draft-spec (verified 2026-08-20)
- A Vocabulary For Expressing AI Usage Preferences (draft-ietf-aipref-vocab-07) — IETF aipref WG (Paul Keller, Open Future; Martin Thomson, Mozilla), draft-spec (verified 2026-08-20)
- Noai and noimageai Tag Adoption: Study and Live Dashboard — Originality.AI, dataset (verified 2026-08-20)
- Meta tags and HTML attributes that Google supports — Google Search Central, vendor-doc (verified 2026-08-21)
- IPTC publishes metadata guidance for AI-generated “synthetic media” — IPTC, spec (verified 2026-08-20)