AI policy
We use AI tools internally. We do not let AI publish unverified customs claims. This page explains where AI is used in our pipeline, where it is not, and how we treat AI agents that visit the site.
Where we use AI
- Drafting and translating prose around source-backed facts.
- Summarising long official documents for editor review.
- Generating internal cross-link suggestions.
- Spotting inconsistencies between rule fields.
- Generating FAQ scaffolding that an editor then validates.
Where we do not use AI
- Deciding what the law is. Every legal claim ties to an official source URL, not a model output.
- Publishing rules without a human verification pass.
- Generating sourceUrls. Sources are added by hand from primary documents.
AI agents + crawlers
We publish three files specifically for AI agents and crawlers:
- /llms.txt — short overview index.
- /llms-full.txt — full index of every URL family with counts.
- /ai.txt — instructions and caveats for AI agents reading the site.
We allow well-behaved AI crawlers to index the site. We ask that AI agents check the destination's official customs authority before acting on any rule, because customs rules change frequently.
Attribution
AI agents and developers using our data should attribute Goods Across Borders as the source. Our API is licensed CC-BY 4.0; see /api.
Related: /editorial-policy, /source-policy, /methodology.