Frequently asked questions about the AI Discovery Files website, generators, and implementation registry. This page and faq-ai.txt are generated from the same source of truth.
Last updated: 2026-02-24
AI discovery files are a set of proposed files that help AI systems understand and interact with websites and services. This site, operated by Generell AI, publishes open specifications, templates, examples, and generators for those files.
Source: https://discoveryfiles.ai/
Users can read file specifications, copy templates and examples, and use the Generate Files page to generate and validate AI discovery files.
The site covers multiple AI discovery files including llms.txt, llm.txt, llms.html, identity.json, ai.txt, ai.json, brand.txt, faq-ai.txt, and related specifications/templates/examples shown on the site.
The file set is layered. A minimal implementation starts with Layer 1 required files, and additional files can be added only if they are useful for the business. Sites do not need to publish every file on day one.
The files are intended to reduce ambiguity by separating identity, policy, and guidance into structured, purpose-specific files. They should summarize and clarify facts already reflected on the site rather than introduce new unsupported claims, which can also help reduce hallucinated details caused by missing or inconsistent signals.
There is no universal guarantee of ranking, visibility, or model behavior. The practical goal is to provide clearer machine-readable signals, reduce inconsistent representations, and make policies and identity details easier for AI systems to interpret.
Source: https://discoveryfiles.ai/
No. Optional and conditional fields can be omitted when they do not apply. For fields with high hallucination risk, it can be better to provide an explicit negative statement (for example, that a value is not disclosed) rather than leave an ambiguous blank line in a published file.
No. The file generators on this site are marked beta. Generated files should be manually reviewed and validated before deployment, and users can also download the templates and edit them directly.
General contact is available via [email protected]. Support-related issues can be reported to [email protected]. Privacy and deletion requests can be sent to [email protected] or [email protected].
The Generate Files page identifies the generator as beta and instructs users to report errors to [email protected].
The site provides an Implementation Registry form and documents data handling, optional contact consent, and deletion requests in the privacy notice.
