Proof

Score by group

What this site currently fails (1)

Not hidden, not rounded away. Each one names why it has not been fixed, because a gap with a stated reason is a decision and a gap without one is an oversight.

• Failingsec-csp

  A real Content-Security-Policy, with the compromises documented

  A real CSP is set with object-src none, base-uri self, frame-ancestors none and a constrained form-action, but script-src allows 'unsafe-inline'. A nonce would force dynamic rendering on every route, which contradicts the static-first requirement. This is published as a known gap rather than hidden.

  Fix: Move to a nonce-based script-src and accept dynamic rendering, or split the deployment

Measurements

Facts that are neither a pass nor a fail. They are published because a number a reader can check is worth more than a badge, and because a check that passes on mechanism can still have an outcome worth stating.

• Sitemap lastmod distribution

  2 distinct days across 195 URLs, largest cluster 175 (90%)

  Above the 25% threshold the integrity report flags. On a site this new the dates are genuine: two consecutive builds produce a byte-identical sitemap, which a build stamp could not. The cluster spreads as chapters are revised on different days.

Requires manual attestation (9)

These cannot be verified by a script. They are listed rather than counted as passes, because counting an unverified check as a pass is how a score stops meaning anything.

• No indexable URL is reachable only from the sitemap

  crawl-orphan-pages

  Run a crawler against production and diff against sitemap.xml

• Answers are self-contained sentences that restate their subject

  aeo-self-contained-sentences

  Read every FAQ answer out of context and rewrite any that starts with we, it, this or that

• Every marketing and programmatic page is prerendered, so a database outage cannot take down the funnel

  perf-static-first

  Prerender the marketing and programmatic routes

• LCP under 1.8 seconds on mobile for every template

  perf-lcp-under-1800

  Run Lighthouse mobile on the deployed site

• No layout shift from the sticky CTA, the cookie banner or web fonts

  perf-no-cls-from-chrome

  Measure CLS on a deployed build

• Every interactive element is keyboard reachable with a visible focus ring

  a11y-keyboard-operable

  Tab through every page and confirm a visible focus ring in visual order

• Search Console queries are segmented by page tier, not read in aggregate

  measure-gsc-by-tier

  Verify Search Console and filter Pages by each prefix

• You can distinguish an indexation problem from a ranking problem

  measure-indexation-vs-ranking

  Compare submitted versus indexed against impressions, per tier

• Watch for over-indexation of a thin tier dragging down the whole domain

  fail-thin-tier-overindexation

  Compare indexed count against pages with impressions, per tier

AI citation checks

Twenty seed queries, fixed at launch, run monthly against three answer engines. The query set never changes, because adding queries mid-series makes the trend meaningless.

A run marked "not run" means no credentials were configured for that engine that month. It is recorded rather than skipped, because "we did not ask" and "it did not cite us" are different facts and conflating them corrupts the series.