Websites that people and their AI assistants can both read.
Stoagen builds AI-friendly websites. Every page is written once, published twice, as a readable page for people and a faithful machine copy for the assistant reading over their shoulder.
Most websites are written for people and merely tolerated by machines. The assistants that more and more readers bring with them get a scraped, mangled, or blocked version of the page, and then answer questions from stale training data instead.
Stoagen sites are written once and published twice. The reader gets a clean, fast, accessible page. Their assistant gets a markdown twin of that same page, plus an appendix of sources and caveats the page itself does not need to show. Neither version can drift from the other, because both are generated from the same file.
See it working
The flagship demonstration is Pueblo SRDA Questions, a guide to the senior services in Pueblo, Colorado, built around the questions its author and a group of friends, all eligible for those services, want answered before they use them. Each program page lays out the public record and the questions that are still open. Paste its link into ChatGPT or Claude and ask something real, like "what help is there for my mother, who lives alone?" The assistant reads the current record, keeps the site's caveats, and answers with the right program and the right phone number.
More on the demo sites page, including what each one demonstrates.
What makes a site AI-friendly
The short version, with the full account on the pattern page:
- One source, two audiences. Pages are authored whole in markdown; the HTML page is the human rendering, and the markdown mirror beside it is the machine record, always a superset, never a contradiction.
- A front door for assistants. One small file describes the whole site to any chatbot, including a single URL that carries the entire corpus in one fetch.
- Honesty as structure. Unverified facts are marked, dates come from revision history rather than someone's memory, and every empirical claim keeps its source.
- Field-tested against real assistants. The pattern's rules come from watching actual fetchers fail, and every rule is enforced by a validator on every build.
Ask your AI about this page
Paste this page's link into ChatGPT, Claude, or any AI assistant and ask your question in your own words. Every page here publishes a machine-readable copy, so your assistant can read the current record directly:
https://stoagen.com/
For example: “What is an AI-friendly website?” · “How do the markdown mirrors work?” · “What has Stoagen built?”
Every page here has a markdown twin; this page's is https://stoagen.com/index.md (also served with .txt appended). The whole site exists as one plain-text file at https://stoagen.com/llms-full.txt, https://stoagen.com/llms.txt describes how the record is organized, and https://stoagen.com/agents/ carries the site's notes for assistants.
Published . Last updated . Times come from this page's revision history and can be checked against it.