CrawlLedger
First-party sources · observed policy states

Know which AI crawler a rule actually controls.

Training, search, user-requested fetching, grounding, and open-web archiving use different robots.txt tokens. This directory combines first-party operator documentation with CrawlLedger's latest observed pilot records.

11documented tokens
6operators
20observed domains
0Content Signal adopters

Current robots.txt policy landscape

Counts come from the latest available robots.txt observation for each domain in the bounded pilot. “No matching group” means the file did not select that token or a wildcard group; it is not proof of deliberate permission.

Crawler tokenDocumented purposeDisallowedPartialAllowedNo match / unverified
GPTBot
OpenAI
Model training7841
OAI-SearchBot
OpenAI
AI search31241
ChatGPT-User
OpenAI
User-requested fetch41141
ClaudeBot
Anthropic
Model training7841
Claude-SearchBot
Anthropic
AI search51041
Claude-User
Anthropic
User-requested fetch51041
Google-Extended
Google
Training and grounding control41141
Applebot-Extended
Apple
Model training6941
PerplexityBot
Perplexity
AI search6941
Perplexity-User
Perplexity
User-requested fetch21341
CCBot
Common Crawl
Open web archive6941

Registry version 2026-08-13 · 20 current robots records represented.

Why the distinctions matter

A single operator may publish separate controls for model training, search discovery, and a human-requested page visit. Blocking one token does not automatically describe the others.

Automatic crawlers

GPTBot, OAI-SearchBot, ClaudeBot, Claude-SearchBot, PerplexityBot, and CCBot identify automatic crawl or indexing functions in their operators' documentation.

User-requested fetchers

ChatGPT-User, Claude-User, and Perplexity-User act after a person asks for content. Some operators explicitly say ordinary robots rules may not apply.

Control tokens

Google-Extended and Applebot-Extended express downstream-use choices but are not separate HTTP fetcher identities.

Evidence boundaries

Every reference page links to the operator's documentation and states when it was reviewed. Observed policy states describe parser output, not authenticated traffic, operator compliance, site ownership, or legal permission.