You finish your answer and wait for the reaction. A nod, a slight lean forward, something. It doesn't come. No shifting in the chair, no small laugh at the part that usually gets one. Just a straight, unmoving face on the other end of the screen, waiting for you to keep going. That's not a bad interviewer. That's not a human interviewer.
Most interview coaching still assumes someone on the other end is steering things along: a nod that says keep going, a raised eyebrow that says wrap it up. That assumption is getting harder to count on. At some companies, AI now conducts the interview directly, not just supports it, and habits built for a room that reacts don't carry over to one that doesn't.
A survey conducted through Pollfish in August 2025 questioned 1,399 full-time U.S. workers in management-level positions with extensive hiring experience. It found that 57% of companies use AI to support hiring. Among companies that use AI in hiring, 34% use it for interviews; among companies using AI for interviews, half allow AI to conduct interviews directly.
The available Middle East data offers a broader indicator of AI maturity in workforce management, not a direct measure of AI interviewing. A report found that 67% of Middle East organizations were still "exploring or just getting started" with using AI to guide talent deployment, while 28% of regional firms reported actively using generative AI.
Here's the part that catches people off guard: companies aren't reporting many negative reactions. The same U.S. survey found that only 4% of companies said they had received negative reactions from candidates about AI's use in the interview process. Worth being precise about what that measures, though: it reflects what companies reported hearing, not how every candidate actually felt.
You'd think that means nothing's really different, right? Except that's backwards. A quiet feedback loop isn't the same as an unchanged interview. Something about the mechanics shifted; it just never had to show up as a complaint to be real.
We think the fix isn't learning to read an AI interviewer better. There's nothing there to read. The fix is building an answer that doesn't need reading, which is the same principle behind PitchPro's Conversation Mode: run the moment enough times that the substance holds up on its own, regardless of reaction.
What Actually Changed
An AI interviewer isn't necessarily just recording you. The same survey found that among companies using AI for interviews, 59% assess tone, language, or body language, and 66% analyze candidate language. It doesn't tell us whether that produces the same live back-and-forth as a person, the confused look that cues you to clarify, or the raised eyebrow that cuts a tangent short.
Many companies do keep a human somewhere in the loop: when AI conducts interviews, 71% always retain oversight, 23% say it depends, and 6% say AI runs independently. The numbers don't say when that oversight happens, so it's safer not to assume anyone's reacting live mid-sentence.
Early stages still carry real weight: among companies using AI in hiring, 39% limit rejection recommendations to the initial stages, while 35% use them at any stage, and only 26% report human oversight for every rejection.
What This Means for Students
Career prep hasn't fully caught up to this yet. Interview coaching, in classrooms and career centers, trains people to read a room, watch engagement, adjust pacing, and build rapport before the harder questions land. None of that transfers when the system isn't signaling anything back.
What does transfer is a narrower, exacting skill: saying the clear version of an answer the first time, without a mid-sentence correction from someone's expression, because it might never come.
A Question Worth Sitting With
If nobody's nodding along, and you cannot count on a follow-up to fill in whatever got left out, how much of what you'd normally lean on in an interview was ever about the content, and how much was someone else quietly helping you land it?
Say the question is familiar: "Walk me through a time you solved a difficult problem." With a person, you might start loose and adjust halfway through based on how they're reacting. If AI is conducting the interview, you can't count on that recalibration. The first version has to be the clear one: the actual problem, the specific action, and the concrete result, stated plainly enough that it doesn't depend on a follow-up to land.
That's not a harder question; it's the same question, minus the reassurance that someone's going to help you find a better version of the answer as you go.
Before You Go
An AI-led interview can still ask familiar questions in a familiar shape. What it's built to notice underneath that shape is what's changed.
Explore how practicing against a room that won't react can become part of interview readiness →