Robot Trust Protocol (RTP v1.0) — Developer Reference

The Protocol

How It Works

Place a robots-trust.json file at the well-known URI. AI agents fetch it before interacting with your site to understand your AI interaction policy.

1 — Place the file

Host the file at:

URL

https://yourdomain.com/.well-known/robots-trust.json

2 — AI agents read it

Any AI crawler or agent checks this URL to get your policy before crawling, scraping, or training on your content.

Reference agent (Node.js)

async function checkTrust(domain) {
  const url = `https://${domain}
    /.well-known/robots-trust.json`;
  const r = await fetch(url);
  if (!r.ok) return null;
  return await r.json();
}

Minimum Required

Minimal Valid Certificate

Only robot_trust_version is required. The simplest valid file:

Minimal (self-declared, no verification)

{
  "robot_trust_version": "1.0",
  "ai_policy": {
    "agents": "allowed",
    "training": "restricted",
    "scraping": "limited"
  }
}

Full example (verified certificate)

{
  "robot_trust_version": "1.0",
  "protocol": "https://robot-trust.org/protocol",
  "ai_policy": {
    "agents": "allowed",
    "training": "restricted",
    "scraping": "limited"
  },
  "trust_status": {
    "certificate_status": "verified",
    "certificate_issuer": "robot-trust.org",
    "trust_level": "pro"
  },
  "access_points": {
    "preferred_entry": "/ai/",
    "rate_limit_policy": "ai-friendly"
  }
}

Field Reference

ai_policy block

The ai_policy object is the core policy signal. AI agents read this first.

Field	Values	Meaning
agents	`"allowed"` · `"restricted"` · `"denied"`	Whether AI agents may interact with this site
training	`"allowed"` · `"restricted"` · `"denied"`	Whether content may be used for AI model training
scraping	`"allowed"` · `"limited"` · `"denied"`	Whether automated scraping is permitted

trust_status block

Optional if self-declaring. Required for verified certificates from robot-trust.org.

Field	Type	Description
certificate_status	required	`"verified"` · `"pending"` · `"revoked"`
certificate_issuer	optional	Who issued the certificate (e.g. `"robot-trust.org"`)
trust_level	optional	`"basic"` · `"pro"` · `"enterprise"`
issued	optional	ISO 8601 date (`YYYY-MM-DD`)
expires	optional	ISO 8601 date (`YYYY-MM-DD`)

Top-level fields

Field	Type	Description
robot_trust_version	required	Always `"1.0"`
protocol	optional	Link to this page so agents can find the spec
schema	optional	Link to the JSON Schema for validation

Examples

Example Implementations

Basic site (allow all)

{
  "robot_trust_version": "1.0",
  "ai_policy": {
    "agents": "allowed",
    "training": "allowed",
    "scraping": "allowed"
  }
}

No training (content protected)

{
  "robot_trust_version": "1.0",
  "ai_policy": {
    "agents": "allowed",
    "training": "denied",
    "scraping": "limited"
  }
}

Private / restricted

{
  "robot_trust_version": "1.0",
  "ai_policy": {
    "agents": "restricted",
    "training": "denied",
    "scraping": "denied"
  }
}

For AI Developers

Agent Decision Logic

Here is reference logic for how an AI agent should implement support for this protocol:

JavaScript — reference implementation

async function applyRobotTrustPolicy(domain) {
  const url = `https://${domain}/.well-known/robots-trust.json`;

  let policy;
  try {
    const r = await fetch(url, { signal: AbortSignal.timeout(5000) });
    if (!r.ok) return; // No policy — proceed with default behavior
    policy = await r.json();
  } catch {
    return; // Timeout or error — proceed with default behavior
  }

  const ap = policy.ai_policy || {};

  // 1. Agent access
  if (ap.agents === "denied") throw new Error("AI agents not permitted on this site");
  if (ap.agents === "restricted") console.warn("Restricted access — check site rules");

  // 2. Training
  if (ap.training === "denied") markAsNoTraining(domain);
  if (ap.training === "restricted") markAsRestrictedTraining(domain);

  // 3. Scraping
  if (ap.scraping === "denied") throw new Error("Scraping not permitted");
  if (ap.scraping === "limited") applyPoliteDelay(domain);
}

Resources

Robot Trust Protocol