How the score is calculated, and what it does not claim

1. 1. Each criterion earns points. Every criterion declares a point value within its category. Binary checks normally award full credit for a pass, partial credit for a warning and none for a failure. Graduated checks use a measured ratio: a passing check can still earn partial points. The status describes its threshold; the displayed earned and available points determine its contribution. Reports label such a pass Partial credit and list every lost point, by category, under “Why this is not 100”.
2. 2. Categories are ratios, not sums. A category score is points earned divided by points available, over the criteria that could actually be measured. Criteria marked informational, not checked or unable to verify carry zero weight, so an unmeasurable criterion never costs the page points.
3. 3. Sparse categories are withheld. A normal category score is not shown when no weighted criterion was measured, fewer than three weighted criteria were measured, or fewer than half of that category's listed criteria contributed points. It is also withheld when one heuristic supplies at least half of the measured weight with fewer than five measured criteria, or heuristics supply at least 75%. These are conservative SEOLens confidence rules, not Google ranking rules. Performance is also withheld when its evidence is transfer-only, because response-transfer observations cannot substitute for a Field or Lab rendering measurement. A legacy raw score across every category with at least one scored check remains in exports for compatibility and auditability; it is not a second readiness verdict.
4. 4. Primary categories are re-weighted. Search implementation readiness uses Crawlability, Technical, On-Page and Performance when they have sufficient evidence. Content heuristics, Structured Data, Accessibility and Social Metadata remain visible supporting measurements with zero effective weight. Included primary categories are re-normalised proportionally. Confidence is low below 65% primary-model coverage, moderate below 90% or when any included category has moderate evidence, and high otherwise.
5. 5. Eligibility is a separate gate. The implementation value does not override HTTP errors, robots blocking, noindex or a conflicting canonical. A URL with one of those blockers is labelled not scorable for search implementation readiness. If a search exclusion has no observed conflict, the roadmap is withheld until the user confirms whether the exclusion is intentional. When the URL answers 4xx or 5xx, advice about the page itself — its canonical, structured data or social tags — is withheld from the roadmap entirely: Google states an error response is not indexed, so there is nothing yet to optimise. Those findings stay in the criteria list with their evidence.
6. 6. A grade needs most of the weight. Re-normalising makes the score speak only for the categories that had evidence, so the same number means different things at 100% and at 68% of the criterion weight. A readiness score therefore carries a quality band only when at least 90% of the primary design weight was measurable and evidence confidence is high. Between 65% and 90% the number is published as a measured score with its quality grade withheld, without a band or verdict colour; below 65% no readiness number is published at all. Every report states the percentage it rests on beside the gauge. Weighted model coverage and real-user field-data availability are shown separately because one cannot stand in for the other. These limits are SEOLens confidence rules, not Google rules.

Missing evidence changes the basis of comparison. Read category coverage and confidence alongside the technical value, which answers: how much of the sufficiently supported scoring credit did the observed checks earn?

✓Pass

The page meets this criterion.

◐Partial credit

The page clears this criterion’s threshold but earns only part of its points. The missing points are listed under “Why this is not 100”.

!Warning

The page partially meets this criterion.

✕Fail

The page does not meet this criterion.

iInfo

An observation with no pass or fail judgement. Not scored.

–Not checked

Out of scope for this analysis. Excluded from the score.

?Unable to verify

We tried but could not obtain trustworthy data. Excluded from the score rather than estimated.

Documented by the source

These come directly from Google or a standards body. They are requirements or officially published thresholds.

LCP

good ≤ 2.5 s · poor > 4.0 sweb.dev (Google)

INP

good ≤ 200 ms · poor > 500 msweb.dev (Google)

CLS

good ≤ 0.1 · poor > 0.25web.dev (Google)

TTFB target

under 800 msweb.dev (Google)

robots.txt size processed by Google

500 KiBGoogle Search Central

Sitemap limits

50,000 URLs and 50 MB uncompressed per fileGoogle Search Central

Core Web Vitals evaluation point

75th percentile of real user visitsweb.dev (Google)

One vocabulary, platform-specific features

The generator creates standard Schema.org JSON-LD, which can be read by any consumer that supports that vocabulary. Google, Bing, Yandex and other crawlers decide independently which types, properties and search experiences they support. A valid Schema.org graph is therefore not a promise of a Google rich result, a Bing enhancement, a Yandex appearance, indexing or ranking.

SEOLens applies documented Google feature rules only where Google publishes a verifiable requirement set. It does not invent parallel eligibility rules for engines whose public documentation does not define an equivalent test. Validate the vocabulary first, then inspect the published URL with each target engine. Bing URL Inspection requires the URL to belong to a site added to Bing Webmaster Tools.

Every declared type and property is compared against the published Schema.org vocabulary — 939 types and 1,538 properties from the release retrieved 2026-09-23. A term that does not exist in it, such as nmae for name, is reported as a fault, because a consumer ignores what it does not recognise rather than reporting it. A real property used on a type it is not defined for is reported without being failed: Schema.org publishes its domains as guidance and states the vocabulary is extensible. This is a vocabulary and documented-requirement check, not a replica of Google's Rich Results Test — confirm eligibility there.

The comparison layer answers one question — did an observed signal change between two audits — and it answers nothing else. It does not decide what a signal means: indexability comes from the same chain the report shows, conflicts from the conflict engine, scores from the scoring engine. It reads their output and asks whether it moved.