Glossary

The instructions, standards, and workflows that let Claude Code or Codex work on a real site reliably across sessions.

A website, an AI operating layer, and a data intelligence layer working as one system rather than three separate tools.

The written architecture rules a site carries with it so every change, by any person or AI, follows the same shape.

The accumulated record of every change made to a site over time, with dates.

A group of related sites in a portfolio that share a subject area and cross-reference each other.

AI doing implementation work inside defined standards, with human approval, rather than generating output nobody can maintain.

The warehouse that records search performance, server activity, crawler visits, human traffic, and dated change history in one place.

An open protocol for publishing standardized machine-readable endpoints so AI agents can discover and evaluate sites consistently.

A site whose content lives in structured files rather than a database, making every change readable, diffable, and reversible.

A legitimate automated crawler, including AI systems reading pages to answer questions, as distinct from scrapers and malicious traffic.

A plain-text file at a site's root describing the site for AI systems, similar in spirit to robots.txt.

A structured JSON file a site publishes so AI agents get real data instead of scraping rendered markup.

Build, instrument, observe, decide, improve, measure again. The repeating cycle an AI Website System exists to run.

Applying and verifying a single fix across every connected site in one pass.

Structured data in a page telling machines what the page is about, in a vocabulary they already understand.

A dated record of a change made to a site: what was done, where, and when.

Reading the web server's own request log so every hit is visible, including ones analytics never sees.

Copy stored as data with named fields rather than trapped inside page-builder markup.