How this glossary is written.

Where the content comes from

The initial 25 terms were drafted in a single push on 2026-05-13 using LLM assistance, then audited by a separate research agent for factual claims before publication. Every body section, FAQ, and source citation went through this two-pass process. Specific numbers, dates, and vendor capabilities flagged by the audit were softened or replaced.

The site is not claiming to be primary research. It is a curated, schema-first reference synthesizing publicly documented facts — schema.org vocabulary, arXiv papers, vendor blog posts, search-engine product announcements — into one indexed, AI-citation-friendly resource.

Fact-check rhythm

Every term page shows a Last fact-checked date in its editorial footer. The cadence:

• Every new term batch runs through an independent research-agent audit before commit, with the agent's old → new edit pairs applied verbatim.
• Weekly source-URL audit via scripts/audit-sources.mjs — dead links are replaced or removed, not silently left to rot.
• Status freshness sweeps — the "Status in 2026" sections are revisited monthly against current vendor announcements; dateModified bumps accompany every substantive edit.

Citation status — what those badges mean

Each term page displays four citation badges (ChatGPT, Perplexity, Claude, Copilot) reflecting whether that AI engine has been observed citing the term. Today most badges read untested because the site is new; they will populate as manual probes and (eventually) automated tracking surface real citation events. The count 47× means "we have observed 47 citation events to date for this term on this engine," not "47% of a fixed query set."

Known limits and what they mean for you

• Where a specific number, date, or vendor claim could not be sourced to a public document, we softened it ("modest CTR" vs. "0.5–2% CTR"). If we missed one, that's a bug — open an issue.
• "Status in 2026" sections reflect best-effort synthesis of the current ecosystem. The AI-search landscape changes quickly and some claims may be six weeks stale before a sweep refreshes them. Always cross-check time-sensitive specifics against the cited sources.
• We do not accept paid placement, sponsored definitions, or "vendor preferred" framing. Tool names appear when relevant to illustrate categories, not because of any commercial relationship.

Reporting errors

During the private-development phase (through 2026-06-19 Day 30 review), corrections go via the founder's existing channels. After public launch, this page will list a GitHub Issues link for reader-reported errata. Every accepted correction will result in an updated lastFactChecked stamp on the affected term.

Who runs this

A solo project by an indie founder building schema-first reference assets — see the homepage for project positioning. No team, no ads, no popups, no lead-magnet gates. The newsletter is the only email collection; sign-up is fully voluntary and we never sell or share the list.