What interview accommodations can deaf and hard of hearing candidates request?

The common four are CART captioning, a sign language interpreter, one interviewer speaking at a time instead of a panel, and each question pasted into the chat as it is asked. CART is the strongest option for a technical round, because a captioner writes verbatim and can load a glossary of names and jargon in advance. The employer arranges and pays for CART and for interpreting.

What is CART and how is it different from automatic captions?

CART stands for Communication Access Real-time Translation. A trained human captioner writes what is said verbatim, usually delivered to you as a streaming text link you open in a second browser window. Automatic captions are generated by speech software inside the meeting platform. The practical differences are accuracy on proper nouns and technical terms, which you can prime by sending a glossary beforehand, and cost, since CART is billed hourly to the employer.

Does the employer have to pay for CART at a job interview?

In the United States, CART and interpreting are reasonable accommodations under the ADA, arranged and paid for by the employer, and the cost cannot be pushed onto the candidate. In the UK they are reasonable adjustments under the Equality Act 2010, and the Access to Work scheme has historically funded communication support at interview. Never book a captioner on your own card expecting to be reimbursed.

How far in advance should you request CART for an interview?

Ask at the moment you are being scheduled, and allow several working days. An agency has to find a captioner who is free in that exact slot, and a minimum booking plus a cancellation window measured in days is common. Send a glossary of names and technical terms about two days beforehand, so it is loaded before the session rather than during it.

Are Zoom or Teams auto-captions good enough for a technical interview?

As a backup, yes. As your only access, they are risky for two reasons. Framework names, internal product names and personal names are exactly what a general speech model gets wrong, and those are the words a technical round turns on. Accuracy also collapses when two panellists talk over each other. On a locked-down corporate account an administrator can disable captions entirely, so test the platform beforehand.

Can you use your own transcription app instead of asking for CART?

Use it as well, not instead. A personal transcription tool is a self-help layer that needs nobody's consent, and it is genuinely useful on the morning a booked captioner fails to appear. It is not an accommodation: it creates no record, commits the employer to nothing, and leaves the identical gap in the next hiring process. Send the request, then run your own backup behind it.

Deaf Interview Accommodations: CART vs Captions vs Interpreter

Ask for CART in writing, to the recruiting coordinator, at the moment the interview is being scheduled. Of the four options below it is the only one that is reliably accurate on proper nouns and technical vocabulary, and the only one you can prime with a glossary beforehand. Platform auto-captions are free and need nobody's permission, but they break on exactly the words a technical round turns on. An interpreter is the right ask if a signed language is your first language, and personal on-device transcription is a self-help layer you run for yourself, never a substitute for the accommodation you were entitled to request.

This is practical wording, not legal advice. In the United States the ADA treats CART and interpreting as reasonable accommodations that the employer arranges and pays for, and it is not lawful to push the cost onto the candidate. In the UK these are reasonable adjustments under the Equality Act 2010, and the Access to Work scheme has historically funded communication support at interview even before you are anyone's employee, so check the current terms rather than assuming. Either way, do not book a captioner on your own card and hope to claim it back.

The four options, scored on what actually decides a round

Latency and cost are the columns people ask about. Accuracy on names and jargon is the column that decides whether you answer the question that was asked, and consent is the column that decides whether you have any recourse when it fails.

	Platform auto-captions	CART	Sign language interpreter	On-device transcription
Lag behind the speaker	Roughly one to three seconds, and displayed words rewrite themselves as the model hears more of the phrase	A couple of seconds. Text is committed rather than rewritten, so a correction arrives as a fresh line instead of the sentence changing under you	About a sentence, because the interpreter waits for enough grammar to render it, plus turn-taking overhead when you answer	About a second end to end, arriving as punctuated sentences
Names, acronyms and jargon	Weakest of the four. Framework names, internal product names and the interviewer's own name are precisely what a general speech model gets wrong	Strongest, and the only one you can prime. A captioner loads your glossary into their dictionary before the session starts	Accurate, but unfamiliar technical terms get fingerspelled letter by letter unless you agree signs in advance, which costs seconds each time	Same class of error as platform captions. There is nowhere to submit a glossary
Does the employer have to agree	No. You switch it on yourself and nothing is announced to the room	Yes. This is a formal accommodation request, booked and paid for by them	Yes, same route, and the harder of the two to source at short notice	No. Nothing joins the call and nothing appears in the participant list
Cost, and who carries it	Free, bundled with the meeting platform	Billed hourly to the employer, commonly with a minimum booking and a cancellation window measured in days	Billed hourly to the employer. Long assignments often need two interpreters working in relay, so a full loop costs more than a single round	Your own subscription. See pricing, including a free daily allowance
Lead time to arrange	None, but confirm the feature exists on their account before the day	Several working days is realistic. An agency has to find a captioner free in that exact slot	Longer still, especially for a specific language or a technical specialism	Install and rehearse with it beforehand
Where it sits on screen	Inside the meeting window, competing for space with the shared code editor	Usually a separate streaming text window, which is the more robust setup because it does not depend on their account tier	A pinned video tile you have to watch, which is the real cost during live coding	Your own overlay, excluded from screen capture at the operating system level
Most common way it fails	Absent on the day because an admin policy disabled it, or accuracy collapses the moment two panellists talk over each other	The captioner is booked for one round of a four round loop, or never receives the joining link	The loop overruns and the booking ends before the last session does	Audio routed away from the PC, or a speech model writing fluent text that nobody said
Where it fits best	A short recruiter screen, and as the fallback behind everything else	Technical and behavioural rounds where the exact wording is the thing being scored	Any round, if a signed language is your first language, and behavioural rounds in particular	Rehearsal, and the round where your primary option fails at nine in the morning

The glossary is the reason CART is accurate, so send one

This is the step most candidates skip, and it is the difference between a transcript you can answer from and one you spend the round decoding. Captioners write from a personal dictionary built up over years, and unfamiliar proper nouns are the known weak point. Every agency will accept a prep pack. Most want it at least a day ahead, so it is loaded before the session rather than during it.

Names and roles from the calendar invite. Attendee names as they are spelled, plus job titles. A mangled surname in the transcript is a small thing until you have to use it in a thank-you email.
The company's own vocabulary. Internal product names, team names, anything from the careers page that would not appear in a general dictionary.
The stack, copied verbatim from the job advert. Kubernetes, Kafka, gRPC, Terraform, whatever is listed. Paste it rather than retyping from memory, so the spelling is theirs.
The acronyms in your own resume. You will say them out loud, and the interviewer will repeat them back with a follow-up question attached.
Ask for verbatim CART, not meaning-for-meaning. Condensed services such as TypeWell and C-Print are excellent in a lecture, where the point is the gist. In a system design round the clause that gets condensed away is the constraint the answer is scored on. Put the word verbatim in the request.
Ask whether you get the transcript afterwards. Often you can, and it makes a specific follow-up email trivial to write. Ask rather than assume, because a saved transcript raises a consent question the coordinator has to settle first.

The booking sequence, and where it breaks

At scheduling, name the service and the format

Not "I have hearing loss", which invites a well-meaning guess. "Verbatim CART, delivered as a streaming text link I open in a second window" is a request a coordinator can forward to a vendor without a single follow-up email.

Get the delivery method confirmed in writing

In-meeting captioning depends on the employer's account settings and on whether an admin has enabled third-party captions at all. A separate streaming text window depends on nothing but a browser. If they offer both, take the browser link as primary and treat the in-meeting version as the spare.

Send the glossary two days out

To the coordinator, with an explicit ask to forward it to the captioning agency. Coordinators file attachments and forget them, so name the deadline: "could this reach the captioner by Thursday so it can be loaded in advance".

Join fifteen minutes early for a caption check

Test the link, the font size and the window position while it is still housekeeping. Fifteen minutes is enough to escalate to a phone call if the link is dead. Two minutes is not.

Re-confirm per round, and ask about overrun

The classic failure on a four round loop is coverage for round one only. Ask two questions in one message: is the captioner booked for every session including the gaps, and what happens if the loop runs thirty minutes long.

The request email

Subject: Interview accommodation request, [Role], [date]

Hi [Name],

Thanks for setting this up. I am hard of hearing and I would like to request CART captioning for the interview. Is there an accommodations contact I should copy in, or are you the right person to arrange it?

What I need is verbatim CART, ideally delivered as a streaming text link I can open in a second browser window, since that works regardless of the meeting platform. If the captioner can be booked for all four sessions rather than the first one, that covers the whole loop in one go.

I will send a short glossary of names and technical terms two days beforehand so it can be loaded in advance. Could you confirm the vendor once it is booked? I will join fifteen minutes early to test the link.

Best regards,
[Your name]

Three things that email does deliberately. It names a service rather than a difficulty, so nobody has to research what to buy. It offers the glossary, which reads as helpful and is the single biggest lever on accuracy. And it raises the whole loop up front, before an agency quote has been signed off for one hour only.

Before you rely on platform captions

Auto-captions are worth switching on in every call, including the ones that already have CART, because they cost nothing and cover the seconds where a human misses. Two specifics decide whether they can be your primary. First, the feature can simply be missing: on a locked-down corporate account an administrator can disable captions, so open a test meeting on the exact platform named in the invite and confirm the control is there. Second, they degrade badly on a panel. Two people finishing each other's sentences is the situation every automatic system handles worst, and it is also the situation you most need help with. Find the caption control before the call starts, because hunting through a menu during a live question costs you the question.

Where a personal transcription tool fits, and where it does not

Say this part plainly: a personal transcription tool is a self-help layer, not an accommodation. It is never a substitute for a formally requested CART service, and it should not become your reason for skipping the email above. The request is the thing that changes the process, creates a record, and gives you something to point at when a round is run without access. Software on your own laptop does none of that, and if you quietly use it instead of asking, the next employer's process has exactly the same gap in it.

The honest limits of the on-device option, ours included:

It produces an English transcript. If BSL or ASL is your first language, text in a second language is not equivalent access, and no amount of speed changes that. Ask for the interpreter.
Speech models invent things. Whisper can emit fluent, confident text during silence or noise that nobody said. If a line surprises you, treat it as unverified and ask, rather than answering the sentence you read.
Crosstalk defeats it, exactly as it defeats platform captions. Separating simultaneous speakers is the hard part of the problem and none of the automatic options solve it.
The audio has to come through the computer. Loopback capture reads system playback, so headphones and AirPods are fine, but a phone dial-in is out. If a round moves to a phone line the transcript stops.
Windows 10 and 11 only. No macOS, no Linux, no mobile, no browser extension, so an in-person interview or a MacBook is outside what this can do at all.
Capture exclusion is a privacy property, not an anti-detection one. The overlay is excluded from screen capture at the operating system level, so it does not appear in Zoom, Teams, Meet, OBS or a screenshot. That covers screen sharing and recording only. It does nothing about a webcam, a lockdown browser, eye tracking, a second-device camera, or someone watching you in the room.

Where it does earn a place is the morning the primary option fails, which happens more often than any vendor page admits: the captioner never appears, the streaming link is dead, the caption button is greyed out on their managed account. OfferPilot AI transcribes the interviewer's audio on-device with Whisper and puts a speakable draft on screen in about a second, grounded in your resume, so you have something to read while the coordinator is being paged. Metering counts listening time only, so a round that stalls costs you nothing. The two quieter uses are probably worth more: rehearsing against unfamiliar phrasing from a question list, and building a company bank with the Q&A generator, which doubles as the glossary you send the captioner. The Interview Copilot page has the rest of the mechanics.

If the answer is no, or is silence

Get it in writing, then narrow the ask. A refusal to fund CART is rarely a refusal to do anything at all, and the cheaper adjustments need no vendor: one interviewer speaking at a time rather than a panel, each question also pasted into the chat as it is asked, a written technical prompt instead of a spoken one, the agenda and attendee list in advance, next steps confirmed by email. Ask them to enable the platform's own captions if an admin policy is blocking them, which is a settings change rather than a purchase. And say the sentence that removes the fairness objection before it gets raised: none of this changes the content of the interview or the time you spend answering.

Captions for a remote interview: CART, auto-captions, an interpreter, or on-device transcription