What Is TDM-REP and Why You Should Care

If you’ve been near AI / copyright conversations recently, you’ve seen “TDM-REP” mentioned and probably skipped over it. Time to back-fill. TDM-REP is becoming the legal standard for machine-readable opt-outs from AI training, and EU compliance is now baked in.

What TDM-REP stands for

Text and Data Mining — Reservation of rights Protocol. Catchy, no. But the substance matters: it’s a W3C-recognised, machine-readable format for declaring whether you reserve rights against text and data mining of your content. That covers AI training as a specific subset.

Why it exists

EU Directive 2019/790 (the “Copyright Directive in the Digital Single Market”) created an exception: text and data mining is generally allowed UNLESS the rightsholder reserves their rights in a machine-readable way. That last clause is the key. Reservation has to be machine-readable — a paragraph in your terms of service doesn’t count.

Article 53 of the EU AI Act (in force 2025-2026) extends this: providers of general-purpose AI must comply with TDM opt-outs. They face fines if they ignore properly-expressed reservations.

TDM-REP is the specific machine-readable format the IPA / W3C community has converged on. It’s how you say “I reserve rights” in a way a crawler can parse.

What it looks like

TDM-REP can be expressed two ways: as an HTTP header on every response, or as a tdmrep.json file at a well-known location on your site.

HTTP header form (per-resource):

X-TDM-Reservation: tdm-reservation=1
X-TDM-Policy: https://example.com/tdm-policy.json

Site-wide JSON form (at /.well-known/tdmrep.json):

{
  "tdmRep": [
    {
      "location": "/",
      "tdmReservation": 1,
      "tdmPolicy": "https://example.com/tdm-policy.json"
    }
  ]
}

The value 1 means “rights reserved” (AI training not permitted without a license). The tdmPolicy URL points to a JSON file with your detailed policy — including commercial license terms if applicable.

How AIOX handles it

Automatic. As soon as you connect your domain to AIOX:

• The Suite plugin emits X-TDM-Reservation headers on every response.
• A /.well-known/tdmrep.json file is generated and served from your domain.
• Both reflect your Content Licensing configuration in real-time. Change your config, headers update on next request.
• If you configure a commercial license URL, it’s referenced from the tdm-policy.

Practical implications

• EU compliance: out-of-the-box. You don’t have to read the AI Act.
• Enforceability: TDM-REP gives you legal standing if an AI provider scrapes your content and uses it for training. The reservation must have been clearly expressed.
• AI providers honour it: most legitimate providers (OpenAI, Anthropic, Google) honour TDM-REP today, voluntarily. The EU AI Act makes that voluntary compliance a legal requirement.
• Doesn’t block honest crawlers: TDM-REP is opt-out from training, not from indexing. PerplexityBot and similar can still cite your content; they just can’t train on it.