What we check

A category’s score is the share of its available points earned. A pass earns all of them, a warn earns half, a fail and an error earn none, and a skipped check leaves the denominator rather than counting as a zero. Changing any number here is a versioned event, not a tweak: every score row records the version that produced it, so history can be re-scored instead of overwritten.

We request your page once as each of these, sequentially, one second apart. The Chrome request is the control and everything else is compared against it.

• chromeUser-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/141.0.0.0 Safari/537.36control
• claudebotUser-Agent: ClaudeBot/1.0agent
• gptbotUser-Agent: GPTBot/1.2agent
• perplexityUser-Agent: PerplexityBot/1.0agent
• googleextUser-Agent: Google-Extendedagent

Every client gets the same answer18 points

agent_status_parity

fails when any agent client receives a different status class than the Chrome control

Page text survives without JavaScript11 points

js_dependency_ratio

fails when ratio above 0.7

warns when ratio between 0.4 and 0.7

First byte arrives quickly3 points

raw_fetch_latency

fails when ttfb above 2500 ms

No long redirect chains3 points

redirect_depth

fails when more than 3 hops

robots.txt exists and parses4 points

robots_present

robots.txt does not silently block reading agents8 points

robots_agent_rules

fails when an agent user agent is disallowed while search crawlers are allowed

Sitemap exists and is not empty4 points

sitemap_present

llms.txt tells agents which pages matter5 points

llms_txt_present

A markdown representation is advertised7 points

markdown_alternate

fails when no link rel=alternate type=text/markdown and no Link header equivalent

One h1, a sane heading order, a main landmark6 points

semantic_landmarks

Each page has its own title and description4 points

title_meta_distinct

Server honours Accept: text/markdown3 points

content_negotiation

JSON-LD describes what this is6 points

jsonld_present

Pricing is machine readable5 points

pricing_structured

fails when prices appear only in images or JS-rendered tables

Canonical URL and OpenGraph are complete4 points

canonical_og

An agent manifest or WebMCP endpoint exists5 points

agent_manifest

API docs are two hops from the homepage4 points

api_docs_reachable

Primary forms have labels, names and autocomplete tokens4 points

form_semantics

Docs and pricing are not behind auth or a captcha2 points

no_wall_on_docs

Last-Modified or ETag is sent3 points

cache_headers

Sitemap lastmod values are maintained2 points

sitemap_lastmod_real

fails when every lastmod is identical or in the future

• A 85–100
• B 70–84
• C 55–69
• D 35–54
• F 0–34

• We never work around a block. If your site refuses BotreadyBot/1.0, we record it as refused and display it as refused. No second request under another user agent, no residential proxies, no captcha solving.
• We read at most 6 pages. Sequentially, 1 second apart. We are a diagnostic tool, not a load generator.
• We obey your robots.txt. Including when it is the thing being measured, and CI asserts it.
• We do not scan behind auth. A URL with credentials in it is refused rather than followed.

The catalog above is data, not code: adding, retiring or reweighting a check is an edit to one JSON file, which is why the numbers on this page cannot disagree with the scores on the result pages.