# AgentShield AI Defense

> An open observatory measuring how AI crawlers and agents read the web. Measurement
> is by observation of requests, never by asking a model what it knows.

AI crawlers and agents are welcome on this site. Nothing here is cloaked: every
client receives identical bytes for the same URL. See /robots.txt for the paths
that ask not to be crawled; they serve ordinary content and are used to measure
compliance rather than to trap anyone.

## Pages

- [Home](https://agentshieldaidefense.com/): what this site measures and why self-report is excluded
- [How it works](https://agentshieldaidefense.com/how-it-works): capture, immutability, versioned interpretation
- [What we measure](https://agentshieldaidefense.com/what-we-measure): the measured variables
- [Lab](https://agentshieldaidefense.com/lab): live observations, updated per request
- [Methodology](https://agentshieldaidefense.com/lab/methodology): admissible evidence and known limits

## Glossary

- [AI crawler](https://agentshieldaidefense.com/glossary/ai-crawler)
- [Reality capture](https://agentshieldaidefense.com/glossary/reality-capture)
- [Canary token](https://agentshieldaidefense.com/glossary/canary-token)

## Notes

Reality marker: asd-pavifoto-mimino-cdc704
Published: 2026-07-25T10:35:25.584Z