Blog

Are AI interviews fair? What candidates report and what the research says so far

What candidates worry about, the early evidence on accents, disability, errors and personal data, the findings that point the other way, and what you can do.

· 7 min read

Close-up of a microphone on a stand against a soft, blurred background.
Photo: Robinson Recalde / Unsplash

The short answer: it is not settled

There is no simple yes or no. Candidates report real problems, and some early studies point to real risks. A large field experiment, covered below, found that an AI interviewer did as well as human recruiters in one setting. The candidate experience is often poor, and you have more options than it might seem.

What candidates report

In a May 2026 Greenhouse survey, 38% of job seekers said they had walked away from a hiring process because it included an AI interview, and another 12% said they would. 70% were not clearly told up front that AI would evaluate them. In the UK, 47% had had an AI interview and 30% had walked away because of one, according to the same research. In a LinkedIn poll reported by JobLeads, more than 36% called an AI first interview a dealbreaker, though the poll's sample size was not stated.

For many people the objection is about respect. "I don't want to work for a company if the HR person can't even spend the time to talk to me," one job seeker told Fortune. That is a fair thing to weigh when you decide where you want to work.

Where the concerns are

  • Speech recognition can struggle with accents. Research reported by Information Age found that one video-interview system's speech-to-text had an error rate under 10% for US English speakers, but up to 22% for non-native speakers with other accents. If answers are assessed from a transcript, those errors carry through, and the report warns of the risk of "harm at unprecedented speed and scale".
  • Disabled and neurodivergent candidates can be marked down for things unrelated to the job. An autistic candidate described a system that "kept flagging my eye contact and mannerisms," adding, "I'm not sure how it isn't disability discrimination." Eye contact, facial expression and pace can all work against people this way.
  • Bias in the training data carries into the results. As one widely liked comment put it, "AI doesn't eliminate bias. AI is as biased as the data it is trained on." In the Greenhouse survey, 29% of job seekers wanted proof of a bias audit.
  • Candidates are rarely told what is measured. 39% of the same respondents wanted a clear explanation of what the AI is measuring, and without one you can't prepare fairly or challenge a mistake.
  • The software sometimes fails during the interview itself. A large field experiment recorded technical problems in about 7% of AI interviews. Viral clips make glitches look more common than that, but 7% is still about one interview in fourteen.
  • Your recordings and transcripts are personal data. In one case reported by Information Age, a fast-food chain's hiring chatbot left more than 64 million applicant records reachable behind an admin login of "123456".

Findings that point the other way

Some results complicate the picture. In the field experiment with about 70,000 applicants, 78% chose the AI voice interviewer when they were given the choice, and applicants interviewed by the AI were 12% more likely to receive an offer. It is a working paper, not yet peer-reviewed, and it covers one setting, so it can't show that AI interviews are fair everywhere. Human interviews have biases too, and they are what the AI is being compared with.

The Greenhouse survey points the same way on one measure: perceived age bias was about the same for AI and human interviewers, at around 36%. And 38% of job seekers came away with a more positive view of an employer after a well-run AI interview. How an employer uses the format counts for a lot.

Are AI interviews legit?

Real employers do use them, so the useful question is whether a particular invitation is real. A real one comes from the company or names a platform you can check on the company's careers page. You should never be asked to pay, to install unusual software, or to give bank details or identity numbers at the screening stage. If something feels off, contact the company through its own website before you continue.

What you can do

  • Ask what is assessed, and whether a person reviews your answers before any decision.
  • Ask how long your recording and transcript are kept, who can see them, and how to request deletion.
  • If you have a disability, a speech difference or a technical barrier, ask for an accommodation or another format. There is a guide with example wording.
  • Give the software a fair chance to hear you, with a quiet room, a good microphone, full sentences and an even pace.
  • Save the invitation, and take screenshots if something goes wrong.
  • Decide what counts as a red flag for you. Walking away is a legitimate choice, and so is going ahead.

Practice where nothing is decided

Whatever you conclude about fairness, being used to the format means the conditions trip you up less on the day. A practice round decides nothing, so it is a safe place to find out how you come across. If you want one, you can practice with an AI interviewer on Hmmlo. We publish how we score practice answers, including the limits of those numbers, and what we keep and for how long. To read about the formats first, start with what an AI interview is.

Rehearse it with Nora

Nora is our AI interviewer. Say your answer out loud before it counts.

Already a member? Sign in

Questions people ask

Are AI interviews legit?

Many employers use them. Check that the invitation comes from the company or a platform it names on its own careers page, and never pay or share bank details to be interviewed. If in doubt, contact the company directly.

Are AI interviews safe?

They involve your recorded voice, image and words, which are personal data. Ask how long recordings are kept, who can see them and how to request deletion. Data incidents have happened, so it is reasonable to ask before you start.

Are AI interviews a red flag?

Not necessarily. Some employers use them carefully, with a person reviewing answers and clear explanations. Others don't tell candidates or never follow up. How the employer handles your questions often tells you more than the format.

What do AI interviews look for?

It varies, and many employers don't explain. Most systems work from a transcript of your answers, so content and clarity count most. Some also look at delivery, such as pace. You can ask the employer what is assessed.

Are AI interviews worth it?

That is your call. Some candidates find them efficient and even easier than a human interview; others walk away. If the role is important to you, preparing for the format and asking your questions up front usually makes it more manageable.

More from the blog

Free

Interview notes, now and then

Short, practical emails: one interview question and how to answer it well. No spam, and you can unsubscribe in one click.