What “reviewed” means.
Every guide your agent receives says whether its current version is reviewed or not yet reviewed. This page says exactly what each label means, and what neither one promises.
Last updated 28 September 2026
Reviewed
An AI review agent, working under the owner’s authority, reviewed that exact version of the guide and approved it. The approval is recorded against the version’s content fingerprint in an append-only log.
If the guide changes later, the new version starts again as not yet reviewed. A label never carries over to text it didn’t cover.
“Reviewed” does not mean a person has read the guide.
Not yet reviewed
The current version hasn’t had that review yet. It may contain errors. Treat its numbers, platform or legal requirements, benchmarks and code as a starting point: check them against current official documentation and test before you rely on them.
Your agent receives this same warning with every guide that isn’t reviewed, and search ranks reviewed guides higher.
What every guide has, either way
- Original ASV Labs writing.
- Implementation steps, a verification checklist, its limitations and the sources it relies on.
- Numbers that AI agents checked in September 2026 against the standards and platform documentation the guides cite (such as WCAG 2.2, Core Web Vitals and Apple and Android documentation) and in Chromium lab tests. Wrong values were corrected; values without a firm source are written as judgement defaults.
- Code examples that were checked to parse. They were not run.
What neither label promises
That what your agent builds will look good, meet a standard such as WCAG, or satisfy a legal requirement. Check what you ship. Nothing in Motif is legal advice; see the terms.
How many are reviewed
As of 28 September 2026, 110 of 738 guides are reviewed. The live count is published at /v1/catalog/stats (reviewed_guides).