Accessibility audit — Viadirect voice guidance, Verona airport
Results of the complete audit carried out on 30 September 2026 on the voice guidance for visually impaired people (https://bridge.viadirect.com/guidage-vocal/verone), against RGAA 4.1.2 and WCAG 2.1/2.2 levels A and AA.
RGAA 4.1.2 (French accessibility framework)
Conformance rate: 100% of tested criteria (47 conformant criteria out of 47 decided criteria). 55 criteria are not applicable and 4 remain untested; both categories are excluded from the calculation, as the official RGAA method requires.
Non-conformances
No non-conformance found.
Untested criteria and reason (4)
| Criterion | Tests without a verdict | Reason |
|---|---|---|
| 7.1 — Is each script compatible with assistive technologies where necessary? | 7.1.2 | Conformance of each user-interface component to a reference design pattern: not automated (name and role rendering was verified with the NVDA screen reader, test 7.1.3 conformant). |
| 7.2 — For each script with an alternative, is that alternative relevant? | 7.2.1, 7.2.2 | The application relies entirely on scripts; deciding whether a fallback element is an alternative, and judging its relevance, is not decided by our rules. |
| 7.4 — For each script that initiates a change of context, is the user warned or in control of it? | 7.4.1 | Changes of context initiated by a script (moving to the next screen after a choice, automatic advance of the step counter): whether the user is informed beforehand is a judgement that is not automated. |
| 12.8 — On each web page, is the tab order consistent? | 12.8.2 | The tab order of each screen is measured and consistent (test 12.8.1 conformant); the order after content is inserted by a script (results list, next instruction) is not measured by our rules. |
WCAG 2.1 / 2.2 — levels A and AA
34 conformant criteria, 0 non-conformant, 19 not applicable, 3 not tested. No non-conformance was found; however, level AA conformance cannot be claimed until the 3 untested criteria have been evaluated.
Non-conformances
No non-conformance found.
Untested criteria and reason (3)
| Criterion | Reason |
|---|---|
| 3.3.1 — Error Identification | Error identification: the application has no validated input (a search with no result announces "Aucun lieu trouvé", verified with NVDA); not decided by our rules. |
| 3.3.3 — Error Suggestion | Error suggestion: no automated method in our approach. |
| 3.3.7 — Redundant Entry | Redundant entry: no automated method in our approach. |
Technical audit conditions
- Date: 30 September 2026. Audited version: application files published on 30 September 2026 (app.js?v=38861ed639, engine.js?v=0180d6d9ce, numbers.js?v=ed28b9e8e5, app.css?v=a93a53dfca).
- Exhaustive scope: 12 screens, i.e. every screen of the application. Four are reachable by URL: language choice, starting-point screen in French, English and Spanish. Eight are reached through a scripted journey: search results, category, location choice, destination, summary, and three guidance steps including a level change. Reference trip: Check-in 1-20 to Gates B.
- Environment: Chromium 149 browser driven by Playwright 1.61, 1400 × 900 px window; additional measurements at 320 px wide, with text enlarged to 200%, modified text spacing and forced-colours mode.
- Automated analysis: axe-core 4.9.1 on every screen (violations, passes and incomplete results), plus about forty in-house checks (visible focus, accessible names, touch targets, computed contrast, hidden content, status messages, gestures, etc.).
- Code validation: Nu Html Checker validator run locally on the rendered code of each screen.
- Visual and semantic judgements (relevance of text alternatives, information conveyed by colour, decorative images, link and heading texts): multimodal language model via API, on screenshots and excerpts kept for review.
- Real screen reader: NVDA 2026.2. Element-by-element comparison between the computed accessible name and the speech output under Firefox 151, on the screens reachable by URL. Captures of the guidance screens under Chromium: announcement of search results, instructions read when the application voice is turned off, "Repeat" button.
- Instrumentation set before each screen loads: global keyboard listeners (single-character shortcuts), sounds started without a gesture, gestures and window opening; tab order recorded from the top of the screen; measured contrast of graphic elements (SVG and glyphs). Motion-related features (compass, step counter): review of the source code of the audited version, cross-checked with the captured screens.
- Real-condition tests: complete keyboard-only journey (standard and step-free routes, from search to arrival); 30 minutes without any action on the guidance screen.
- Calculation: an RGAA criterion is conformant only if all its applicable tests are conformant on every screen; a single non-conformant test makes it non-conformant; a criterion with tests still lacking a verdict is "not tested", never conformant. Rate = conformant / (conformant + non-conformant). The rate including model judgements and the mechanically proven rate are both 100%.
- Reasoned assessment: the application is a step-by-step process ("Step 1 of 3" to "Step 3 of 3", then guidance), a particular case provided for by RGAA criterion 12.1. The "Search a place" field is a step in this process, not a site search engine. Criteria 12.1 and 12.5 were therefore judged not applicable, including the tests that had been measured as conformant.
- Limits: no testing with users with disabilities; no VoiceOver (iOS) or TalkBack (Android) testing on a real phone; no on-site testing with real sensors (accuracy of instructions, precision of the compass and step counter); augmented-reality mode and other airports out of scope.
Audit carried out by Accessibility Sentinel — www.accessibility-sentinel.com