Accessibility on the AI Lawyer (UK), and what we do not yet offer in British Sign Language
Dogetlawyer is a UK Legal-Technology platform, not a law firm. We are not regulated by the Solicitors Regulation Authority and we do not provide legal advice — the AI Lawyer gives general legal information and is not a substitute for advice from a qualified solicitor or barrister on your own situation. This page sets out what our accessibility features do today, what is built but not yet available, and what we deliberately do not claim — so you can check us before you use us.
What you can use today
Everything below is deployed and running on the live platform. It sits in one panel on the AI Lawyer chat page, reached from the last button in the chat's left icon rail, or a floating button if that rail is not present — so you are not pushed into a separate 'accessibility mode' with fewer features. The panel closes when you send a message so the reply can use the full width, and reopens from the same button. Our own testing is server-side: it confirms the panel loads, the routes respond and audio is returned. It does not confirm that microphone capture works on your particular device, browser or assistive technology — we have not completed end-to-end testing on real hardware, so please test the voice features before you depend on them. If a feature is not described in the present tense on this page, assume we have not built it.
Push-to-talk dictation
Hold the microphone button and speak. On desktop browsers that support it, a live caption preview appears while you talk — labelled as a preview, because it is your browser guessing. On phones and tablets there is no live preview: we switch it off there so it cannot compete with the recorder for the microphone. Either way, a full transcript from our third-party speech-to-text provider arrives when you let go. Dictation is billed against your account's credits in the same way as a typed chat message.
An editable review box
The push-to-talk transcript lands in an editable box with chips for dates, amounts, numbers and email addresses. You correct what was misheard, then choose Insert, Send or Discard.
Answers read aloud
Listen buttons and an auto-read option speak the AI Lawyer's replies through our server-side text-to-speech route, which passes the text to a third-party voice provider. If that route is unavailable, playback falls back to your browser's own voice — it will sound different, and we would rather it kept working than stopped.
A hands-free loop
Opt in and the panel keeps listening. It has its own consent step, and each turn is then sent automatically unless you cancel within a four-second countdown — so this mode has no read-before-send step. There is a ten-turn cap, and Escape or leaving the window stops it.
Text size, contrast and motion
Font size, contrast and reduced-motion preferences live in the same panel, so you set them where you are reading instead of hunting through a settings page.
Dictation in the CRM as well
If your account has access to the CRM agent, that chat has its own microphone beside Send. The transcript lands in the message box and the script never presses Send for you.
British Sign Language: built, empty, not yet available
We have built the BSL clip library and it is empty. No clip has been published and no signed content is available yet. That is the honest position, and we would rather publish it than let you assume otherwise. A clip is only ever served when three things are true at once: it is approved, it has media attached, and its approval has not expired. Anything marked as unreviewed AI output can never be approved, attaching new media resets approval to unapproved, and approval lapses after twelve months and must be renewed. Until a clip passes a named, qualified BSL review, the BSL section shows an honest empty state rather than something that resembles signing. We will not describe the library as available until a qualified BSL reviewer and a Deaf co-design partner have been engaged, and neither has been engaged yet. British Sign Language is a language in its own right, with its own grammar — it is not English on the hands, and it is not American Sign Language, which we do not offer either. We do not offer live sign-language translation, sign recognition, lip-reading or camera gesture control; text-to-BSL generation and camera gesture commands are staged in our specification and have not been built.
How a spoken question becomes a sent message
The design assumption is that speech recognition will get things wrong, especially with names, case numbers and amounts. So in push-to-talk dictation — the default — nothing you say reaches the AI Lawyer until you have read it and agreed to it. Two modes work differently and say so before you start: the opt-in hands-free loop sends each turn automatically unless you cancel within four seconds, and realtime spoken conversation (administrator accounts only at present) streams your speech straight to the model with no review step.
Hold to talk
Press and hold the microphone — a sustained press for as long as you are speaking. The panel shows plainly that the microphone is on.
Watch the preview, on desktop
On desktop browsers that support it, a live caption preview appears as you speak, labelled a preview because it is a browser guess rather than the final transcript. On phones and tablets there is no preview at all.
Read the transcript
On release, the audio is transcribed by our third-party speech-to-text provider and the text lands in an editable review box.
Check the chips
Dates, amounts, numbers and email addresses are chipped, so the details most often misheard are the easiest to check.
Insert, Send or Discard
You decide. Nothing is sent automatically from the push-to-talk review box.
If this is you
Accessibility is not one need. These are the specific things this build does for specific people, with the gaps named rather than glossed over.
Deaf and hard of hearing
Nothing in the AI Lawyer depends on hearing: it is a text chat, and voice is an optional layer on top. We are not going to pretend that settles it. BSL is a language in its own right, and written English is a second language for many BSL users — text-only access is not equal access. The BSL clip library is built but contains no clips, because none has yet passed a named, qualified BSL review, so today the honest answer is English text. And the AI Lawyer is software: AI-generated legal information, not advice from a human solicitor.
Blind and low vision
Replies can be read aloud on demand or automatically, with a fallback to your browser's own voice if the server route is unavailable, and you can raise font size and contrast in the same panel. We have not yet tested the panel with screen readers, so treat its keyboard and screen-reader behaviour as unverified until we publish that testing — read-aloud and the text controls are what we are claiming here.
Speech differences
Because the push-to-talk transcript arrives in an editable box and nothing is sent from it without you choosing Send, a misrecognised word costs you an edit rather than a wrong question. Typing stays available at every point, and typing carries no voice charge.
Motor differences
Push-to-talk removes typing — but note that it needs a sustained press for as long as you are speaking, which does not suit everyone. If holding a button is difficult, the opt-in hands-free loop removes the button entirely: it keeps listening between turns, with a four-second cancellable countdown, a ten-turn cap and Escape to stop. There is no single-tap dictation yet, and no camera gesture control.
Welsh, Irish, Scottish Gaelic and Bengali speakers
This one is about language rather than disability, but it comes up in the same place: we do not offer dictation in Welsh, Irish, Scottish Gaelic or Bengali. Speaking those languages will not produce a usable transcript — please type instead. What is enforced is a permitted-script filter on voice transcripts: Latin and Bengali characters pass, and the scripts we have listed as unexpected are blocked from being displayed, saved or inserted into your chat.
What happens to your voice, and the limits we enforce in code
Audio you dictate goes from your browser to our server and is then passed to a third-party speech-to-text provider to be transcribed; the audio file is deleted from our servers once the text comes back. Read-aloud works the same way — the text of the reply is sent to a third-party text-to-speech provider and the audio is streamed back to you — and realtime spoken conversation streams audio from your browser to that provider directly. Our privacy notice and our How we use AI page name the processors we use. If you would rather no third party processed your voice, type instead: every voice feature on this page is optional and the chat works without it. On cost: dictation is charged against your account's usage allowance in the same way as a typed message, read-aloud is capped per user per day at 20,000 characters by default and then switches to your browser's own voice rather than failing, and realtime spoken conversation carries its own daily time limit. If voice is how you access the service rather than a convenience, tell us before you buy and we will confirm what your plan allows. On correctness: realtime speech recognition is pinned to English at the recogniser, while push-to-talk transcription still detects language on its own, so we back it with a permitted-script filter — CJK, kana, Hangul, Thai, Cyrillic, Arabic, Hebrew and Devanagari are never displayed, never saved and never put in the chat box. We added that after finding language auto-detection returning English speech as Thai and Chinese text, and it applies to voice transcripts only: anything you type yourself is untouched. Only one feature may hold the microphone at a time; starting a realtime conversation stops push-to-talk first, because two consumers of one microphone starve each other and produce garbled half-sentences. The accessibility panel loads only on the pages it belongs to: the AI Lawyer chat page carries its configuration, the CRM page does not, and the public site carries neither.
What we have not tested, and what we will not claim
We publish no WCAG or EN 301 549 conformance claim, because no formal conformance audit has been carried out. We have not tested this build with a screen reader — not NVDA, JAWS, VoiceOver or TalkBack — and we have not tested with refreshable braille displays, so we make no claim about screen-reader or deafblind access. Our verification so far is server-side checks of rendered pages: they confirm the panel renders, the routes respond and text-to-speech returns audio, and they say nothing about microphone capture on your device, how the panel is announced, or whether every control can be reached from the keyboard alone. An end-to-end test on real hardware is still outstanding on our side. A data protection impact assessment covering voice capture, transcription and read-aloud, including the third-party providers involved, has not been completed; until it is signed off we recommend you do not dictate case names, client identities or other confidential detail — type those instead — and we will publish the date of completion on this page. Realtime spoken conversation exists in the product but is currently enabled only for our own administrator accounts while we settle what it costs to run at scale; ordinary accounts get push-to-talk and read-aloud, and see a plain message rather than a broken button. The directory of independent legal professionals that the panel links to is not yet populated, so we cannot currently refer you to a human adviser. If an accessibility feature does not work for you, or you need something on this site in a different format, write to accessibility@dogetlawyer.com and tell us what failed; we do not yet publish a response-time target, we are not able to provide interpreters, and we do not offer a signed or telephone advice line.
For law firms and advice services assessing us
If you are evaluating this for a service where accessibility forms part of the contract, take the claims above literally and test them. The whole accessibility layer has a single off switch and the CRM voice feature has its own — but both are platform-level settings that we operate, and today they are all-or-nothing rather than per-firm, so switching them is a request to us rather than a control in your own admin. You can confirm the panel does not load on pages outside the AI Lawyer chat. You can open the BSL section and see the empty state for yourself. We do not hold a WCAG or EN 301 549 conformance certificate, so we cannot supply one; what we can give you is a feature-by-feature list you can test yourself, and the limits above in writing. Buying or deploying Dogetlawyer does not make your organisation compliant with the Equality Act 2010: the duty to make reasonable adjustments is anticipatory and rests with you as the service provider, and you cannot discharge it by pointing at a supplier's feature list. Nothing on this page is legal advice about your own duties; take that from your own advisers.
Frequently asked questions
Does Dogetlawyer support British Sign Language yet?
Can the AI Lawyer turn my typed question into British Sign Language?
Can I use the AI Lawyer without typing?
Will the AI Lawyer read answers aloud for blind and low-vision users?
What happens to my voice recording after I dictate?
What if it mishears me and sends the wrong thing?
Does it use my camera to read signs or lip movements?
Is Dogetlawyer WCAG 2.2 AA compliant?
Can I dictate in Welsh, Irish, Scottish Gaelic or Bengali?
Can our firm turn these accessibility features off?
See what works today
BSL is not available yet — the clip library is built but empty. Working today: push-to-talk dictation, an editable transcript, and answers read aloud.
Explore more:
Last reviewed 3 August 2026. Dogetlawyer AI Ltd provides technology and AI-generated legal information and document support. It is not a law firm and does not replace advice from an appropriately qualified professional. Check the jurisdiction, date, source and your own circumstances before relying on an output. England and Wales, Scotland and Northern Ireland are separate legal systems.