Employers are leaning on AI to get through the flood of applications, and candidates are pushing back hard. In a 2026 Greenhouse survey of nearly 3,000 job seekers, 38% said they had walked away from a hiring process because it required an AI interview, and 34% came away with a worse opinion of the company. And, the candidates most comfortable walking away are often the most sought after. Here is what these interviews look like, and the 6 things candidates hate most about them.
1. A machine makes the decision
On Mercor, every role requires an AI interview, which you must pass in order to reach a human round. Obviously, candidates prefer a real recruiter who simply watches your answer and makes the call. More tools are now putting a machine in front of the recruiter to make the first decision. Being filtered by software, with no human backstop, is what people resent most, and is garnering attention from lawmakers.
2. You're not told the AI is judging you
In the survey, 70% of candidates weren't told upfront that AI would evaluate them, and 1 in 5 only realized that once the interview had started. Finding out mid-answer that a model, rather than a recruiter, is scoring you is a fast way to lose someone's trust before the conversation has even begun.
3. One take, a countdown, and no do-over
Many one-way tools give you a single attempt on a timer. On HireVue, most setups allow no retakes, so if you stumble in the first 10 seconds, that's the version that gets judged. And it's not just for one question; it can be for several, up to 8 in some tools.
4. The bot watches you for signs of cheating
AI interviews increasingly monitor the candidate as closely as the answers. Tools like Apriora flag tab-switching and background voices as possible cheating, and that kind of active monitoring is one of the top reasons people abandon these interviews. Being treated as a suspect before you have said a word sets a hostile tone.
5. It grades how you look and sound
When an AI processes your video, it can pick up your accent, your appearance, and your expressions, which have nothing to do with whether you can do the job. HireVue built its early product on facial-expression analysis and dropped it in 2021 after an FTC complaint, and most tools still won't say for sure what they analyze. Candidates notice, and they assume the worst.
6. No one will say what it measures or where your video goes
Candidates are left guessing on two fronts: what the AI is scoring, and what happens to the recording of their face and voice afterward. Few vendors state plainly whether your video trains their model or whether you can delete it. People say they want a straightforward explanation of what the AI measures and confidence that a human reviews its output.
Hiring platform comparison
We've compiled a comparison summary of AI interview platforms. The columns are candidates care about most, whether a human is guaranteed to watch, whether the recording trains someone's model, whether an algorithm is grading their face and voice, are the ones nearly every vendor leaves blank or answers badly. These complaints are structural, built into how the category works rather than any single product, which is also why they are fixable. Candidates prefer screens that that keeps a human in the loop, states plainly what the AI does and does not do, gives people more than one take, and lets them own and delete their recording.
