A failure-seeking assessment of Prepsted’s public website, followed by documented remediation and targeted regression testing.
Pass for WCAG 2.2 Level AA within the stated scope.
The initial live assessment found two WCAG failures. It also recorded two conformance risks and five highest-standard advisories. Every recorded item was corrected in the release that includes this report. Targeted post-remediation checks found no remaining applicable Level A or AA failure in the corrected templates.
The audit began without treating previous remediation work or third-party scores as evidence. Every public page and published article was checked, including responsive states, not-found states and the public sign-in and password-recovery pages. The contact form’s error and success journeys were exercised, including one authorised live submission.
The initial release failed because three article category labels narrowly missed minimum text contrast and a decorative trust-page image was given a spoken alternative. Both defects arose from shared presentation or template patterns and were corrected at source.
WCAG conformance is not a percentage: every applicable Level A and AA success criterion must be satisfied on each complete page and responsive variation in scope. A single failure prevents a full conformance claim. See W3C’s explanation of conformance.
WCAG 1.4.3 Contrast (Minimum), Level AA — corrected.
Small category text on all three article pages used #df0b5d over a pale gradient. Its endpoint ratios were 4.38:1 and 4.47:1, below the required 4.5:1. Article-header category text now uses #d20a55, producing approximately 4.86:1 and 4.96:1 at the two endpoints.
WCAG 1.1.1 Non-text Content, Level A — corrected.
The trust hero’s decorative toolbox image was exposed as “Toolbox icon”. It now has a null alternative (alt=""), so assistive technology ignores it while the adjacent page content supplies all meaningful information.
Related to WCAG 1.3.1 Info and Relationships — corrected.
The trust features previously followed the page H1 with six unexplained H3 headings. A real H2 now names the feature group and each feature remains correctly nested beneath it.
Governance and assurance risk — corrected.
The accessibility statement now links to this evidence, records the initial findings and states that the corrected templates completed targeted regression checks. Its conformance wording therefore describes the corrected release rather than the failed pre-remediation state.
Highest-standard advisory — corrected.
The public sign-in and forgotten-password pages no longer rely solely on browser-default focus styling. Links, fields and buttons within the login card now use a robust two-colour indicator: a 3-pixel white outline with a 6-pixel dark outer ring.
Highest-standard advisory — corrected.
Homepage component styles were moved from the body into the shared stylesheet. The service-status time element now has a valid initial machine-readable timestamp. Genuine sections on the homepage, trust page and news page now have identifying headings; the schools-page quotation is no longer incorrectly marked up as an unnamed section.
Highest-standard advisory — corrected.
Muted public-site text and authentication button colours were adjusted to provide a practical buffer above the normative minimum, reducing the chance that small future rendering or colour changes create a borderline failure.
Discoverability advisory — corrected.
The sitemap now lists published articles, legal pages, service status, accessibility, this audit report, sign-in and password recovery as well as the principal marketing pages.
Usability advisory — corrected.
Failed sign-in and password-recovery feedback now receives programmatic focus, uses appropriate alert or status semantics and retains a non-sensitive email address after an error. Password values continue to be cleared.
The reported Skynet Technologies result was reproduced: 50.87%, “Semi Compliant”, with 54 homepage failures. Its free scan covered only the homepage and mixed Level A, AA and AAA rules in one percentage.
All 54 flags were inspected. The 48 contrast entries were false positives when compared with computed text and background colours, including both endpoints of gradients. Four empty-alternative images had no title and were correctly decorative. The header logo was correctly ignored inside a home link that already had an accessible name. The remaining heading flag asserted that the H1 must precede the main element, which WCAG does not require. Ten entries were truncated duplicates. The checker did not identify either of the two genuine failures found elsewhere in the site.
W3C explains that automated tools cannot determine conformance on their own. The third-party score was therefore treated as a source of leads, not as an accessibility verdict.
The assessment covered the homepage, How it works, For schools, For trusts, News and insights, About, Contact, Frequently asked questions, Service status, Privacy, Cookies, Accessibility, Terms, sign-in, forgotten password, all three published article pages, the 404 page, an unknown route and a missing-article state.
Password-protected administration and application pages after sign-in, third-party websites and unpublished content were outside this public-site audit.
Within the tested scope, pages had descriptive titles and valid language declarations, one main landmark and one H1. No duplicate identifiers, positive tab indices, unnamed controls, unlabelled form fields, broken ARIA references or remaining heading-level jumps were detected.
The first public-site focus target is the visible-on-focus “Skip to main content” link. The shared focus indicator remains visible on light, dark and gradient surfaces. Mobile navigation exposes its expanded state and label, closes with Escape and returns focus. At 320 CSS pixels, all tested pages reflowed without page-level horizontal scrolling or content crossing the viewport boundary.
Contact-form errors use a focused summary with links to invalid fields, field-specific descriptions and retained values. Successful submission uses a focused status message. FAQ questions use native disclosure controls. Service status provides textual states, a polite atomic announcement, a timeout fallback and a no-JavaScript explanation.
Every tested image has an alternative attribute. Decorative images use null alternatives and informative images have contextual text alternatives. No public audio, video, iframe, PDF, office document, ZIP or other downloadable content was present, so media-caption and downloadable-document requirements were not applicable.
The assessment combined browser accessibility-tree inspection, complete DOM and computed-style analysis, responsive viewport testing, focus-state and interaction checks, live form and service-status observation, source review, HTTP checks and HTML validation.
The browser test surface did not provide reliable native hardware Tab and Shift+Tab traversal, operating-system forced-colour switching or an actual VoiceOver or NVDA speech session. These areas were corroborated through DOM focus order, absence of positive tab indices, individual keyboard focus states, interaction behaviour, the browser accessibility tree and the applied preference-media rules. This is a self-assessment, not an independent legal certification or a substitute for testing with disabled users.
No known WCAG 2.2 Level A or AA failure remains in the corrected public release. After upload, Prepsted should confirm that the revised files are served without stale caching and repeat a short keyboard and assistive-technology smoke test. Accessibility should be rechecked whenever public content, templates or functionality materially change.