You apply for a dev role. A day later you get an email: "Congrats, you're moving to the next step!" You click the link expecting to book a call with a recruiter or an engineer.

Instead there's nobody. Just your webcam, a question on the screen, a countdown timer and a little red recording dot.

Welcome to the AI interview. If it caught you off guard, you're not the only one.

This is way more common than people think

Greenhouse surveyed almost 3,000 job seekers for its 2026 Candidate AI Interview Report. 63% said they had already been interviewed by an AI, up 13 points in just six months.

And most people aren't loving it. The Guardian ran a story where UK job hunters called these interviews "awkward and humiliating." People talked about getting cut off when they paused to think and never knowing if a real human watched their answers.

The Greenhouse numbers back that up. 70% said they weren't clearly told upfront that an AI would evaluate them. 38% have walked away from a hiring process because of one. And of the people who finished one, 51% never heard back.

As devs we know exactly how these systems work under the hood, which honestly makes it weirder. But they're not going away, so let's talk about how to do well in one.

What an AI interview usually looks like

The common version: you open a link, a question appears, you get maybe 30 seconds to think, then one to three minutes to record your answer. Repeat.

Under the hood it varies. Some just record video for a recruiter to watch later. Others transcribe your answer and score it against the role, basically matching what you said to what the job needs. Some are live voice agents that ask follow up questions. A few dev roles even pair this with an automated coding test.

You're usually not told which one you're getting.

Before you start

Read the instructions. How many questions, how long per answer, can you retry. People skip this screen and panic when the timer starts.

Use the practice question. Most platforms have one that doesn't count. Test your camera, mic and lighting there.

Set up properly. Quiet room, light in front of you, camera at eye level. Close Slack, Discord and every other notification source. Seriously, all of them.

Check your connection. Plug in if you can. Freezing mid answer is rough.

Ask for an accommodation if you need one. If you're neurodivergent, have a disability or need more time, you can ask the employer for an adjustment or a human interview instead. In Ontario, employers are required to accommodate disabilities in hiring. It's a normal request.

How to answer when nobody's nodding back

In a normal interview you read the other person's face. Here you get nothing, so structure matters way more.

Use STAR. Situation, Task, Action, Result. It keeps you from rambling when there's nobody to interrupt you. For technical stories, make the Action part specific: what you changed, which tool, why.

Lead with the answer. Don't spend 40 seconds on background. Say your main point first, then explain. If the clock runs out, the important part is already recorded.

Use the language from the job post. If the posting mentions "REST APIs," "PostgreSQL" and "code review," and you have done those things, say them naturally. If something is scoring a transcript, it's probably matching against the role.

Give real numbers. "I cut our CI build time from 25 minutes to 6 by caching dependencies and splitting the test suite" beats "I'm good at optimizing pipelines," for an AI and for any engineer who watches it later.

Explain technical stuff simply. The first viewer, human or AI, might not be an engineer. Say what the problem was and why it mattered before getting into how you solved it.

Look at the lens, not your own face. A sticky note next to the camera helps.

Things that trip people up

Reading a script word for word. It's obvious on video and sounds flat. Keep a few bullet notes, not full sentences.

Long silences. Some systems cut you off or count them against you. If you need a second, say "Good question, let me think of the best example" instead of going quiet.

And please don't run an LLM in another tab to answer for you. A lot of platforms watch for tab switching or reading off screen. Getting flagged is way worse than a slightly imperfect answer, and you'll have to back up everything you say in the next round anyway.

After the interview

If you don't hear back within a week or so, email the recruiter and ask about next steps. You can also politely ask whether a person will review your recording. With 51% of people never hearing back, a short follow up helps you stand out.

Practice before the real thing

The best prep is answering out loud, on camera, with a timer. Record yourself on your phone answering a few behavioral questions, like a tough bug you fixed, a disagreement in code review or a project you're proud of. Watch it back. It feels awkward once, then it gets way easier.

If you want the questions you're likely to get, ResumeStart's interview prep gives you role specific questions with sample answers based on your resume and the job. Practice them out loud with a timer before the real thing.

The bottom line

AI interviews feel strange because they are strange. Nobody naturally talks to a webcam with no one on the other side.

But they're becoming a normal step in hiring, dev roles included. Set up properly, keep answers structured, use the job post's language, give real examples with numbers and practice on camera a few times. That's what gets you past the bot and into a room with actual engineers.