Methodology

1. Crawl

The pages in scope are discovered from your sitemap and by following internal links from the starting address, with pages nearest that address tested first, up to the page cap of your order. Each page is loaded in Chromium through Playwright at a desktop viewport of 1280 by 800 pixels; the first ten pages are also loaded at a mobile viewport of 375 by 812 pixels. Query-string tracking parameters are ignored so the same page is not tested twice. Because you have commissioned the review of your own site, robots.txt is recorded but not obeyed; requests are limited to one per second.

2. Engines

Every page is tested with two independent rule-based engines: axe-core (all WCAG 2.0, 2.1 and 2.2 Level A and AA rules, with best-practice rules recorded separately and never counted toward conformance) and HTML_CodeSniffer using its WCAG 2 AA standard. Where both engines report the same defect on the same element, it is counted once. Engine warnings that ask for a manual check are recorded as review prompts, not as defects.

3. Probes

Our own checks cover what rule engines test weakly or not at all:

• A sequential keyboard focus walk of up to 120 stops, recording focus order, whether each focused element shows a visible indicator, and whether focus becomes trapped.
• Reflow at a 320 CSS pixel viewport, detecting content that requires horizontal scrolling.
• Landmark and heading structure, a main landmark, a skip link, page language and title.
• Media inventory: video and audio elements, caption tracks, autoplay and controls.
• Viewport zoom restrictions, timed refreshes, live regions, forms and frames.
• Pointer target size against the 24 by 24 pixel minimum of WCAG 2.2, excluding inline text links.

4. Documents

PDF documents linked from tested pages, up to 25, are inspected for a tag structure, a real text layer, a title and a language. Untagged or image-only documents are reported with a recommended action.

5. Assessment

The consolidated evidence for each success criterion (counts, representative instances, probe results, review prompts, the document inspection and a site inventory) is assessed by an expert system built on a frontier language model, which assigns a conformance level in the terms defined by ITI’s VPAT 2.5 (Supports, Partially Supports, Does Not Support, Not Applicable, Not Evaluated), writes the remarks, the executive summary and the remediation plan, and produces a review memo that lists every criterion it could not settle and why.

6. Human review

A person reads the review memo and the flagged criteria, opens the pages in question where needed, and releases, re-crawls or amends before anything reaches you. Every document records the document identifier, the report date, the pages tested and the tooling used.

Standards and versions

Assessment is against WCAG 2.2 Levels A and AA, which includes every WCAG 2.1 criterion referenced by the U.S. Department of Justice’s Title II rule and the EN 301 549 web clauses, with Level AAA reported where evidence exists. Conformance reports use the ITI VPAT 2.5Rev structure in the WCAG, Section 508 or EU edition. Criteria data is generated from the W3C’s published machine-readable WCAG files, not typed by hand.