Methodology
How the score works
Every claim on this page is generated from the same check definitions the scanner runs. Nothing here is aspirational copy.
Scoring
Each of the 14 checks below carries a fixed weight. A passing check earns its full weight; a failing check earns zero. The total available weight is 110. Score = earned weight divided by 110, scaled to 100 and rounded to the nearest whole number.
A check flagged "excluded" (see Partial results below) contributes to neither the numerator nor the denominator: it is treated as not part of that scan, not as a fail.
Grade bands
- A+: score 90 to 100
- A: score 80 to 89
- B: score 65 to 79
- C: score 50 to 64
- D: score 35 to 49
- F: score 0 to 34
The 14 checks, grouped
Content
agents.md
agents.md is the primary self-description file AI agents look for when deciding how to interact with a site.
llms.txt
llms.txt is the emerging convention for pointing LLMs at the most relevant content on a site.
llms-full.txt
llms-full.txt provides an extended, unabridged version of llms.txt for agents that can handle more context.
Content negotiation
Agents that request text/markdown should receive a lightweight markdown response instead of full HTML.
Structured data (JSON-LD)
JSON-LD structured data gives agents a machine-readable summary of what an organization or page is.
sitemap.xml
A sitemap helps agents discover the full set of pages worth crawling or indexing.
Meta hygiene
Title, meta description, and Open Graph tags give agents (and the humans they act for) a reliable summary of the page.
Content readability
If the homepage is almost entirely markup/script with little extractable text, agents that don't execute JS see an empty page.
Trust
robots.txt AI access
robots.txt controls whether AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended) can access your site at all.
MCP server card
An MCP server card at a well-known location lets agents discover tool/action interfaces automatically.
.well-known/verified.md
Publishing /.well-known/verified.md is the strongest first-party signal that a site actively participates in agent-readiness self-attestation.
HTTPS + redirect
Agents should reach the site over HTTPS, and plain HTTP should redirect there so credentials and payloads are never sent in the clear.
Bot access
If the homepage returns a 403 or a challenge page to a scanner-style user agent, real agents likely get blocked too.
Payments
x402 payments (experimental)
x402 is an emerging standard for agent-native micropayments over HTTP 402. Early support signals forward-looking infrastructure.
Partial results (bot-blocked sites)
If a site's bot-protection layer blocks or challenges our scanner's homepage request, 4 checks that depend on reading that same homepage (Content negotiation, Structured data (JSON-LD), Meta hygiene, Content readability) cannot run honestly. Rather than count them as fails and unfairly tank the score of a site that is otherwise fine, we mark them "excluded" and the report as partial. The real score may be higher than what is shown; the bot-access check itself still fails, since if our scanner is blocked, real AI agents likely are too.
Re-verification policy
Certified domains are re-scanned on a weekly cron, but only once their last verification is more than 28 days old, so the practical cadence is monthly. A re-scan counts as a pass only above 80 (grade A). 2 consecutive failed re-verifications lapse the certification, not one, so a single transient error or brief regression does not silently kill it. A lapsed certification's badge switches to a neutral, not-current state; the domain owner is emailed on every failing re-verification, not only once the badge is lost.
The score guarantee
Done-for-you carries a score guarantee, measured by this same public methodology: 95 or higher on our scan, or your money back. On platforms with a documented ceiling the guarantee adjusts to that ceiling. See how we work for the day-by-day and the exact terms.
What the score does not claim
A score is the result of automated, heuristic checks at the time of a scan. It is not a guarantee that a site is accurate, safe, or trustworthy, only that it passed or failed specific technical signals. It is not an endorsement of the business, its content, or its offers, and it is not a guarantee of future performance. See Terms for the full legal language.