Community DNS Dataset Tracks Observed Access to Generative AI Services

Summary

AI Service Indicators is a community-maintained list of DNS hostnames for generative AI applications and APIs. It maps hostname patterns to services, vendors, categories, and roles such as chat applications, public APIs, coding agents, authentication, CDNs, telemetry, and websites. App, API, and agent roles count as usage, while portal, web, auth, CDN, telemetry, download, and other roles do not. The project warns that DNS access shows only that a device contacted a service; it does not prove that a prompt was submitted, identify the person or model involved, or show what the service did. It recommends lowercasing hostnames, removing trailing dots, applying first-match wildcard rules, and classifying the client-requested name rather than a CNAME target. Results should be reported by role and counted as distinct devices per service over a day or week, because caching, retries, background refresh, and prefetching distort query counts. DNS measurement misses off-network devices, external encrypted-DNS resolvers, locally run models, AI features embedded in other software, and vendor-backend calls. Services such as OpenRouter, Azure AI Foundry, Bedrock, and Vertex AI may front multiple models that DNS cannot distinguish. The repository documents regex and Splunk lookup implementations, provides a helper for filtering high-volume DNS indexes, welcomes corrections, and requires its validation script to pass; it is licensed under CC BY 4.0.