How the grade is computed

Every scan runs the same transparent checklist. Each check passes, warns, or fails; checks are weighted inside nine categories; the categories combine into one 0-100 score and a letter grade. Here is the whole thing.

90-100

80-89

70-79

60-69

below 60

How scoring works

Each check is pass (full credit), warn (half credit), or fail (no credit). Checks that do not apply are marked N/A and left out of the math entirely.
Within a category, each check carries a weight by impact. The category score is the weighted percentage of credit it earns.
Categories combine into the overall score using the weights below. If a whole category is N/A (for example, no real-user performance data), it is dropped and the rest are re-weighted, so missing data never counts against you.
The overall grade is weighted for the page type - general, content/article, marketing/landing, or API/app - which we auto-detect (and you can switch in any report). Page type also sets which checks apply: a landing page isn't marked down for skipping an FAQ, outbound citations, or a visible date the way an article is. Those go N/A instead of counting against it.

Categories & signals

Crawlability

16% of grade

Can agents even get in?

Page allows indexing (no noindex in meta robots or the X-Robots-Tag header)
robots.txt exists and is valid
Major AI crawlers are not blocked (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot, Bytespider, Applebot-Extended)
sitemap.xml exists and is referenced in robots.txt
Primary content is in the server HTML, not injected by JavaScript
Returns HTTP 200 with no excessive redirect chain
No login or consent wall hiding the main content
Real content, not a soft 404 (a 200-status error page)
Small enough to read in full (no agent truncation risk)
No cross-host redirect that loses agents

Understanding

16% of grade

Can they parse the meaning?

Descriptive <title> tag
Meaningful meta description
Exactly one <h1> with a logical heading order
Semantic landmarks (header, nav, main, article, footer) instead of div soup
Structured data: JSON-LD present, valid, and using high-value schema.org types
Open Graph + Twitter card tags
Canonical link
html lang attribute
Image alt-text coverage
Form controls are labeled
Main content starts near the top, not buried under boilerplate
Content is not hidden behind tabs or interaction

Answer Readiness

10% of grade

Is it formatted as the answer? (AEO)

Leads with a direct, self-contained answer (not a long warm-up)
Question-style headings that match how people prompt AI
Scannable lists, numbered steps, and comparison tables

Citability

12% of grade

Will AI actually quote you?

Content is extractable as clean, selectable text
A clear, quotable answer near the top of the page
FAQ or Q&A structured content
Author / publisher credibility signals
Descriptive internal linking

Trust & Authority

12% of grade

Do AI engines see a credible source? (E-E-A-T)

Served securely over HTTPS (HSTS a plus)
Identity established as an entity: Organization/Person schema with sameAs links
About, contact, and privacy/terms pages present
Cites authoritative external sources on claim-bearing pages (.gov, .edu, Wikipedia, DOIs, standards bodies like W3C/IETF, official product/developer docs)

AI engines lean hard on E-E-A-T and entity recognition when choosing who to cite, so these signals carry real weight.

Content Depth

8% of grade

Enough substance to be worth citing?

Substantive content length
Organized into clearly-headed sections
Includes supporting media (images, figures, video)
Concrete data and specifics, not just generic claims

Freshness

6% of grade

Is it current?

Visible published / updated date
datePublished / dateModified in structured data
sitemap.xml reports <lastmod> dates

AI systems discount content they cannot date - visible and structured dates keep you on the right side of the recency cliff.

AI-Native

8% of grade

Forward-looking signals most sites have not adopted yet.

/llms.txt (and optional /llms-full.txt)
A clean markdown or plain-text version of the page
AI-bot rules or Content-Signal directives in robots.txt
MCP / agent-discovery file under /.well-known/
Discoverable OpenAPI / API catalog
OAuth / OIDC discovery metadata

These are opportunities, not failures. Absence is normal today and weighted lightly, but adopting them puts you ahead as agents start acting on sites.

Performance

12% of grade

Fast enough to get rendered?

LCP - Largest Contentful Paint (loading)
CLS - Cumulative Layout Shift (visual stability)
INP - Interaction to Next Paint (responsiveness)

Measured with real Chrome user data from the Chrome UX Report (CrUX), 75th percentile, phone data preferred. If Google has no field data for the site yet, Performance is marked N/A and dropped from the grade instead of counting against you - and you can run an on-demand PageSpeed (Lighthouse) lab test, shown separately as synthetic data that never changes the grade.

What we don't do

We don't crawl your whole site. One URL per scan, on purpose.

We don't execute JavaScript - and neither do most AI crawlers. That gap is exactly what we measure.

We don't store your report or your IP. We do keep the URLs you submit (with a timestamp) to understand usage and improve the tool - see the privacy policy.

Scan your site