ARCH

Methodology and limits

Version 2026.10-2

What we read

Your homepage, robots.txt, sitemap, llms.txt and at most one product or room page. At most 10 requests, 30 seconds and 2 MB per page. We identify ourselves as ARCH-ReadinessCheck and read only public pages.

How it’s scored

Each check has a weight from 1 to 3. Verified earns full weight, Partially verified half, Not found none. Not verified and Not applicable are excluded. Product, price and booking checks apply only when we find a public store or booking path. An overall score is shown only when at least 60% of the applicable weight could be verified.

What it can’t tell you

It’s based on publicly verifiable website signals, not a complete assessment of agent capabilities. It doesn’t see real agent traffic or private agent conversations, signed-in checkout or booking flows, product feeds or marketplace listings. It reads one sample page, not your whole site, and doesn’t predict how any AI assistant ranks or recommends a business.

Safety

Only public domains on standard ports. Private, local and reserved network addresses are refused at validation and again at connection time. Redirects are re-checked, scans are rate-limited and results expire after 24 hours.

Scored checks

CheckWeight
Your site answers automated visitors3
Secure connection1
AI assistants and answer agents are allowed3
Content is readable without running scripts2
Structured data is published2
Your organization is clearly identified1
Products are described in structured data3
Products carry identifiers2
Pages have clear titles and summaries1
Prices are machine-readable3
Availability is machine-readable2
Shipping and returns are machine-readable1
A sitemap lists your pages2
Product pages are discoverable2
A path to purchase or booking is visible2
llms.txt guidance file1