The pre-audit in three points
Targeted scope
Up to 20 selected pages, or up to 20 (Free) or 50 (Pro) discovered pages on the same domain, analyzed at 1366×900 desktop and 390×844 mobile sizes.
Automated detection
axe-core in Chromium, plus a focus visibility check and a 24×24 pixel touch target check.
Human validation
An optional RGAA grid and a checklist guide the checks that automation cannot decide.
01 — SCOPE
How pages are analyzed
The result describes a specific point in time, on the pages and states that were actually loaded during the scan.
- In targeted mode, only submitted URLs are analyzed. In discovery mode, Odeet follows internal links three levels deep from the home page, supplemented by the sitemap, up to 20 pages (Free) or 50 pages (Pro) on the same domain. When the site has more, Odeet alternates between sections and page templates to cover each part of the site rather than just one.
- Each URL is opened in Chromium in two contexts: desktop and mobile. Redirects and content rendered at scan time are included.
- axe-core reports violations detectable in the DOM. Odeet adds affected selectors, HTML excerpts, failure messages and remediation guidance.
- Two complementary checks feed the same report: detectable keyboard focus visibility on desktop and minimum touch target size on mobile.
- An HTTP, network or navigation error remains attached to its context so an incomplete scope stays visible.
02 — TECHNICAL SCORE
A score from 0 to 100, where 100 is the best signal
The score helps prioritize work and track change across a comparable scope. It is calculated from axe-core impacts and Odeet’s complementary checks.
Penalty per rule and context
Critical
+4 per occurrence
Serious
+2 per occurrence
Moderate
+1 per occurrence
Minor
+1 per occurrence
The first occurrence applies the main weight. Up to 8 additional occurrences per rule and context are counted, so repetition matters without overwhelming the whole score.
Reading levels
Final calculation
- 1A penalty is calculated separately for every successfully loaded page × viewport context.
- 2The overall penalty is the rounded average of successful contexts, capped at 100.
- 3Score = 100 − average penalty. Adding pages therefore does not mechanically lower the score when their quality is comparable.
- 4Only rules mapped to a WCAG success criterion count toward the score. axe-core best practices (landmark structure, heading order, skip link…) are still reported, but separately and outside the score: they match no WCAG success criterion. This applies to audits created since it was introduced (calculation method v2); earlier audits keep their original method and their score does not change.
How scan errors are handled
When at least one context succeeds, failed contexts are reported but excluded from the average: the score must never be read without checking the failure count. If no context can be analyzed, the score is 0.
03 — REFERENCES
Automated detection never establishes conformance by itself
Odeet maps some findings to WCAG references and an RGAA grid to prepare an auditor’s work. This mapping supports analysis; it is not a certification.
- WCAG and RGAA mappings use the official correspondences published by axe-core for each rule, never an approximate guess.
- When a rule mapped to a criterion fails, that criterion can be prefilled as “non-compliant” with an automated source. A best practice never prefills a criterion: it remains a hint for the auditor.
- The absence of a violation never automatically produces a “compliant” status: the criterion remains subject to human verification.
- A complete absence of frames, media, tables or forms can propose the corresponding topics as “not applicable”. This proposal must be confirmed against the real scope.
- A manual auditor decision always takes precedence over automated prefilling.
- When used, the RGAA rate follows the formula compliant criteria ÷ (compliant + non-compliant). It remains provisional while applicable criteria are untested.
The built-in grid contains the 106 criteria from RGAA 4.1.2, the official version currently in force. The 4.1.2 errata do not invalidate earlier audits.
04 — LIMITS
What still requires human evaluation
A page can receive a good score and still be difficult or impossible for some people to use.
- The relevance of text alternatives, labels, reading order or error messages.
- Complete keyboard navigation, focus traps, dynamic components and end-to-end business journeys.
- Real use with screen readers, voice control, zoom, user preferences and different assistive technologies.
- Content behind authentication, consent, an interaction or a state that the automated scenario did not open.
- Sample representativeness, exemptions, applicable legal scope and the drafting of an accessibility statement.
- Stability over time: the report is a dated snapshot and can become obsolete after a site change.
05 — PROPER USE
How to use the report without overreading the score
- 01
Check the scope
Review the URLs, both viewports and any failed pages before making a comparison.
- 02
Confirm findings
Reproduce priority issues and remove false positives in the product’s real context.
- 03
Test journeys
Add keyboard, zoom and assistive technology tests on a representative sample.
- 04
Communicate carefully
Present the score as a pre-audit indicator. Reserve conformance conclusions for a complete audit based on the current official reference.
A useful pre-audit starts with a clear scope.
Choose relevant pages, review scan errors, then turn confirmed findings into an action plan.
Further reading
- Automated accessibility testing: what it finds and what it missesWhat a robot finds, what it misses and the checks to do by hand.
- RGAA vs WCAG: what’s the difference?WCAG sets the rules; the French RGAA is the method used to check them.
- Test a page’s accessibilityFree test on desktop and mobile, no account needed.
- Audits with your agency’s brandingA PDF report with your logo, ready for your clients.