Accessibility Statement

1. The standard we work to#

We work to the Web Content Accessibility Guidelines (WCAG) 2.1 at level AA, the level the European standard EN 301 549 uses for web content.

2. In every conversation with an organisation's avatar#

• Captions of what the avatar says, on by default, in three sizes, with the option to show your own words too.
• A written transcript beside the conversation. Screen readers find it as a log, and it never talks over the avatar.
• Keyboard access to every control, a link that skips straight to the session controls, and visible focus. The single-key shortcuts (M for the microphone, C for captions, T for the transcript) can be switched off.
• Screen readers are told when the conversation is connecting, connected, lost or back, and when the browser holds the sound back.
• A label for the whole conversation that says the avatar is an AI, with text for screen readers, and a short notice when it starts.
• Motion follows the reduced-motion setting of your device.
• Website visitors and job candidates can type the whole conversation instead of speaking, for example when a microphone is blocked or missing.

3. Options an organisation can turn on#

These are off unless the organisation that runs the avatar turns them on. Where they are on, each person changes them for themselves in the conversation's Accessibility menu, and the choices are kept on their own device only.

• Bigger captions, white or yellow text on solid black, and captions over the picture for those who prefer it. By default captions sit off the picture and the controls under it, so the avatar's face and hands stay in view.
• A mode for deaf and hard-of-hearing people: captions stay on and large, with your own words and the transcript open.
• Easy-read: the avatar uses short sentences and plain words. Job interviews never change the questions, so easy-read is not offered to candidates.
• Slower speech: the avatar speaks more slowly, with pauses.
• A still picture instead of the moving face, with the voice and captions carrying on.
• Screen-reader announcements of who is speaking and when it is your turn.
• For website visitors, a copy of their own conversation's words, as text or as a captions file, made in their browser.
• On learner pages, a reading font designed for low-vision readers.

4. What doesn't work well yet#

• There is no audio description of the avatar's video. What the avatar says is in the captions and the transcript; its expressions and gestures are not described.
• There is no sign-language avatar or interpreter.
• Typing into a spoken conversation, as your own turn, isn't available yet. It is coming soon; until then a person who can't speak can use the typed conversation instead, where the organisation offers one.
• Your own words in the captions and transcript come from speech recognition and can contain mistakes.
• How much slower speech slows down depends on the avatar's voice.
• A live conversation has a maximum length, and it ends after a long silence unless you answer the question that asks whether you are still there.
• Some parts of the console, such as charts and some older settings pages, have not yet been checked to the same standard as live conversations.
• We cannot check the documents and content that organisations add themselves.

5. How we check#

• Automated tests check the markup of the live conversation's screens for missing names, labels and roles, and the contrast of our colours, including captions over the brightest video.
• Where an organisation's own colours are used, we adjust them so that text stays readable.
• We review new features against WCAG 2.1 level AA before they are released.

6. Tell us about a problem#

If something doesn't work for you, or you need information in another format, email hello@zentiqai.com with the page you were on and what happened. We read every message and reply by email. If your question is about an organisation's own avatar or content, we pass it to that organisation.