AI interview bias is the risk that software evaluating your interview rates you lower for reasons that have nothing to do with the job, such as your accent, a speech difference, a disability or patterns in the data the tool learned from. The concern is official: the U.S. Justice Department's guidance on AI in hiring warns that facial and voice analysis can screen out qualified people with autism or speech impairments, and New York City requires a bias audit before employers use automated hiring tools. You can't audit the software yourself, but you can read the notice, ask for accommodations and make your evidence hard to misread.
This guide sticks to what's documented, with every source checked in October 2026, and to what you can do about it. It isn't legal advice. Take the assessment yourself and follow the employer's rules — use CoPilot Interview to practice beforehand.
Where AI interview bias can come from
Bias doesn't need anyone to intend it. It can enter at any step between your voice and your score.
| Where it can enter | What can go wrong | What helps you |
|---|---|---|
| Speech-to-text | Accents, stutters, speech differences or unfamiliar names are transcribed wrongly, and a transcript-based score only sees the wrong words | Clear audio, short full sentences, key terms said slowly; an accommodation if transcription is a real barrier |
| Training data | Tools that compare applicants with current successful employees can repeat past patterns; the Justice Department notes that people with disabilities may be missing from those groups | Nothing you can change, which is why audits and human review matter |
| Voice and face analysis | Qualified people with autism or speech impairments may be screened out, per the Justice Department. California's rules say an automated system that analyzes tone of voice or facial expressions may discriminate based on race, national origin, gender or disability (regulation text) | Ask what's analyzed, and request another format if it puts you at a disadvantage |
| Timers and formats | Strict time limits can disadvantage people who need more processing time or use assistive technology | Ask for extra time or a removed timer. HireVue, for example, says it provides accommodations including allowing retakes and removing timers |
| Proxy data | Details such as ZIP codes can stand in for protected traits | Illinois bars employers from using ZIP codes as a proxy for protected classes; check the rules where you live |
| Vague criteria | Ratings for "fit" or "energy" leave room for impressions instead of evidence | Answer with evidence for the competency the job lists, so there's something concrete to rate |
Most of these risks sit with the tool and the employer, not with you. What you control is the quality of the evidence you give and whether you speak up before the interview. For how the scoring step works, read how AI evaluates interview transcripts.
What bias audits do and don't cover
Bias audits are the main legal check on AI hiring tools in the U.S. right now, and they're narrower than many people assume.
- New York City, Local Law 144. Employers and employment agencies that use an automated employment decision tool for jobs in the city must have an independent auditor complete a bias audit within a year before use. At a minimum, the audit calculates selection or scoring rates and impact ratios across sex, race/ethnicity and intersectional categories, and the employer must publish a summary, usually on the employment section of its website (city FAQ). Disability, age and accent aren't among those required categories, and the city's FAQ says the law doesn't require any specific action based on the results, though anti-discrimination laws still apply.
- Illinois. An employer that relies solely on AI analysis of video interviews to decide who gets an in-person interview must report the race and ethnicity of applicants who are and aren't selected, and of those hired, to the state each year. The state then reports whether the data shows racial bias (820 ILCS 42).
- Vendor testing. Some vendors describe their own checks. HireVue, for example, says its data and algorithms are audited to check for bias. A vendor's audit covers its product in general, not necessarily the way your employer set it up.
How to use this: if the job is in New York City, search the employer's careers page for "bias audit" or "automated employment decision tool" to find the published summary. Our guide to AI interview laws and candidate rights covers the notices you're owed in each place.
Accents, speech differences and disability: your options
The clearest protections here come from disability law, not AI law. The Justice Department's guidance, dated May 12, 2022 and still published on ADA.gov as of October 2026, says the ADA requires employers to provide reasonable accommodations during the hiring process unless that would cause undue hardship, and that hiring tests must measure job skills, not a disability. It also says employers should tell applicants what technology is used and how they'll be evaluated, and make sure that asking for an accommodation doesn't hurt an applicant's chances.
Things you can reasonably ask for, depending on the barrier:
- Extra time per question, or the timer removed.
- A retake if a technical problem or a speech difference spoiled a recording.
- An accessible version, for example one that works with your screen reader.
- A different format, such as an interview with a person or written answers.
- Confirmation that a person reviews your recording or transcript, not only the score.
An accent on its own isn't a disability, so a disability accommodation may not be the right route. You can still ask whether the tool supports your language and whether a person reviews transcripts, and California's rules name national origin among the characteristics that tone-of-voice analysis may discriminate against. For language-specific practice, see our tips for non-native English speakers, and for wording an accommodation request, our accessibility guide for candidates.
What you can do before the interview
- Read every notice. In New York City, candidates who live in the city must get notice 10 business days before an automated tool is used, including the qualifications it assesses and how to request a reasonable accommodation. In Illinois, you must be told how video AI works and asked for consent.
- Ask what's evaluated. One line by email is enough: "Does the AI score my spoken words only, or also my voice or video?" Our guide to whether AI interviews analyze tone of voice shows what vendors say.
- Ask early and in writing if you need an accommodation. Describe the barrier and the change that would fix it. A request before you start is easier to grant than a redo afterward. Our guide to asking for a human interviewer instead of AI includes an email template.
- Test your setup. A headset microphone and a quiet room reduce transcription errors for everyone, and they matter more if you have an accent or speak softly.
- Practice out loud. Our free AI interview assistant includes a Mock Interview mode that scores each spoken answer out of 10 with what to fix and a model answer, so you can see whether your evidence comes through before the real interview. The free plan gives you 15 minutes every day.
Make your evidence hard to misread
You can't control the model, but you can control how much clear evidence it gets. Short sentences with one fact each survive transcription better than long ones, and they leave less room for anyone's impressions, human or machine.
Before (easy to mis-transcribe and easy to misjudge): "So we did the K8s migration thing and it, um, it made the incidents go way down, like a lot, and the team was happy with that, I think."
After (Example, we wrote it): "I led our move to Kubernetes, the container platform, over three months. I wrote the migration plan, moved twelve services one at a time and set up alerts for each one. Production incidents fell from about eight a month to two. The main lesson: move one service first and learn from it before you move the rest."
Why it holds up: the acronym is spelled out once, numbers are said as whole phrases, every sentence starts with what you did, and the result is a measurable change. If one word does get misheard, the sentences around it still carry the evidence. The STAR story builder helps you draft stories in this shape.
If you think an AI score was unfair
- Ask for a human review. Many employers have a person review AI results; ask whether one did in your case. HireVue, for example, says humans make the final hiring decision.
- Ask for information about the tool. In New York City, employers must post how to request information about the tool's data retention, the type of data collected and its source, and answer written requests within 30 days.
- Know what's coming in Colorado. From January 1, 2027, Colorado residents who get an adverse decision materially influenced by covered automated technology can ask how to correct factually incorrect personal data and request human review and reconsideration, to the extent commercially reasonable.
- Know where complaints go. The Justice Department's guidance says you can file a charge of discrimination with the EEOC if you believe your rights were violated because of a disability. In New York City, discrimination claims go to the city's Commission on Human Rights.
- Keep records. Save the invitation, any notice, the dates and every email.
Rules differ by place and change often, so check the law where you live. For the bigger picture, our guide to AI interviewers explains formats, and how to reduce interview bias shows what fair processes look like from the hiring side.
See how your answers read to a scorer
Answer real interview questions out loud and get scored feedback, so your evidence is clear before an AI or a recruiter hears it.
Try the free AI interview demoFAQ
Are AI interviews biased?
They can be. Regulators have flagged specific risks, including voice and facial analysis screening out people with autism or speech impairments, and tools trained on past hires repeating old patterns. Whether a particular tool is biased depends on how it was built, tested and set up, which is why some laws require audits.
Can AI interviews discriminate against accents?
Speech-to-text can mishear accents, and a transcript-based score only sees the words that were captured. California's rules also warn that tone-of-voice analysis may discriminate based on national origin. Use a good microphone, speak in short full sentences and ask whether a person reviews the transcript.
What is an AI bias audit in hiring?
In New York City, it is an independent evaluation of an automated hiring tool that, at a minimum, compares selection or scoring rates across sex, race/ethnicity and intersectional categories. Employers must have one done within a year before using the tool and publish a summary.
Can I ask for an accommodation in an AI interview?
Yes. The U.S. Justice Department's guidance says the ADA requires employers to provide reasonable accommodations during hiring unless that would cause undue hardship. Examples include extra time, a removed timer, an accessible version or a different format. Ask early and in writing.
Does a human review AI interview results?
Often, but it depends on the employer. HireVue, for example, says humans make the final hiring decision. If it matters to you, ask the recruiter whether a person reviews every result before a decision is made.
What should I do if I think an AI interview treated me unfairly?
Ask the employer for a human review and for information about the tool, keep copies of every notice and email, and check the rules where you live. Disability discrimination charges can be filed with the EEOC, and in New York City discrimination claims go to the city's Commission on Human Rights.