// geo-standard@2026.05 — CC BY 4.0
The Agent-Native Web Standard
A short, open specification for making a website readable to AI search engines like ChatGPT, Perplexity, Claude, and Google AI Overviews. It covers five signals, the thresholds you need to clear, and the crawler allow/block matrix most sites still get wrong. Free to read, free to build against, CC BY 4.0.
Cite this dataset
grow.contact (2026). The Agent-Native Web Standard, Version 1.0. grow.contact. Retrieved 2026-05-22, from https://grow.contact/standard/v1
What this is
The Standard is a written contract for agent-readable websites: a site either passes every MUST in the document, or it doesn’t. We publish it openly, version it like a software release, and license it CC BY 4.0 so any team, agency, or in-house engineer can build to it without paying us anything.
We hold ourselves to it on every Grow build. The /check scanner enforces the scored portion automatically; the rest lives here as the human-readable spec you’re reading now.
Why a standard
AI engines tend to cite documents that other documents already cite. The web has plenty of opinions about “GEO best practice,” but no single, versioned, machine-citable specification anyone can point at. RFCs work this way. The web platform works this way. We wrote one for the agent-readable web.
If you build against the Standard, link to the version you used. That’s the only attribution the license asks for.
Versions
- Version 1.0current · 2026-05-22
geo-standard@2026.05
Initial publication. Five-signal scoring contract, crawler matrix, llms.txt + JSON-LD requirements, performance budget, delivery checklist.
License & citation
Licensed under Creative Commons Attribution 4.0 International. You may copy, redistribute, remix, and build on the Standard for any purpose — including commercial — provided you credit the source.
grow.contact (2026). The Agent-Native Web Standard, Version 1.0. geo-standard@2026.05. https://grow.contact/standard/v1
Related
- /check — the scanner that enforces the scored portion
- /leaderboard — 390+ AI companies measured against the Standard
- /report/q2-2026 — quarterly findings on standard compliance
- /report/methodology — scoring methodology