How to prepare for an AI video interview

You open the link expecting a scheduled call with an actual person, and instead you get a countdown timer, a webcam preview, and a list of questions that appear one at a time with nobody on the other end to react to what you say. That's not a technical glitch. It's the modern one way AI video interview, and for most job seekers it's now the norm at the first stage of hiring, not the exception.

The position worth stating first: this format is genuinely disorienting the first time you meet it, mostly because almost nobody explains it to you in advance. The data on that transparency gap is worth knowing before you go in, because it changes how you should prepare, and it means the awkwardness you feel is a documented, widely shared experience rather than something wrong with you specifically.

How common AI interviews actually are now

Sixty three percent of job seekers have now faced an AI interview, a jump of 13 percentage points in just six months, according to Greenhouse's May 2026 candidate survey of 2,950 candidates across the United States, United Kingdom, Germany, Ireland and Australia. That pace of change matters: a format that was unusual eighteen months ago is now something most active job seekers will meet at least once during a search.

The shift looks similar from the hiring side. The Interview Guys' 2026 report, published June 2026, cites survey data showing 96 percent of US hiring professionals now use AI in at least some part of their recruiting process, and 64 percent of organisations already using AI apply it specifically to recruiting, interviewing and hiring decisions, not just resume sorting. (A separate Resume Genius survey of hiring managers found a lower 79 percent figure for "automated at least some part of hiring," a different question, phrased differently, put to a different sample, so the two numbers aren't directly comparable even though both describe AI's spread into hiring.) The same report notes that by the end of 2025, 81 percent of interviewers at large technology employers said they believed candidates were using AI to cheat during remote interviews, a detail worth knowing because it explains some of what an AI interview tool is actually built to detect, alongside evaluating your answers.

Put together, this is no longer an edge case worth dismissing. If you're actively applying right now, the odds that your next first round interview is a recorded, one way AI assessment rather than a live conversation are close to a coin flip and rising.

What an AI interview is actually assessing

An AI video interview typically shows you a written or recorded prompt, gives you a short prep window, then records your answer with no live interviewer present to follow up, redirect, or react. The recording is later scored, and vendors that publish details on how their tools are used generally describe that score as a starting signal for a human reviewer rather than an automatic pass or fail decision on its own, though practice varies by employer and platform.

Most platforms in current use score some combination of:

• The content of your answer. Whether it actually addresses the question asked, with a specific example rather than a generic claim, is the single largest factor across most systems, and it's the part within your full control.
• Structure and clarity. Answers that have a clear beginning, middle and end, and that don't wander, tend to score and read better, both to the algorithm and to whoever reviews the recording afterward.
• Delivery signals. Pace, filler words, and pauses are commonly tracked. This doesn't mean you need to sound like a newsreader, it means rehearsing out loud, not just in your head, helps.
• In some systems, facial and vocal signals. This is the most contested layer, and increasingly a legally scrutinised one. Where it's used, it's typically a secondary signal behind the content of your actual answer, not the primary basis for a decision.

None of this is a reason to try to perform for the algorithm instead of answering the actual question. A strong, specific, honest answer is still the strongest input into every one of the signals above.

The two layers most systems are built around

Most AI interview tools are commonly described, by vendors and hiring teams, as working in two stages rather than one. The first stage is automated scoring against the criteria above, which usually happens within minutes of you finishing. The second stage, where a given employer's process still has one, is a human recruiter or hiring manager reviewing either your full recording or a shortlist the automated pass produced. Whether that second stage reliably happens is a separate, better documented question: the 51 percent of candidates in Greenhouse's survey who said they never heard back after completing an AI interview is real evidence that a human stage doesn't always follow quickly, or in some cases at all, which is a current shortfall in how the process is run, not something you're doing wrong.

Why nobody tells you upfront, and what that means for you

Seventy percent of candidates in the Greenhouse survey said they were never clearly told upfront that AI would be evaluating them, and 21 percent said they only realised once the interview had already started. That gap is the single most consistent complaint in the data, more than objections to AI itself: candidates aren't primarily upset that a machine is involved, they're upset that nobody said so.

Practically, this means you should assume any interview described as "recorded," "on demand," "one way," or scheduled through a platform you don't recognise as a live video tool is likely AI evaluated, even if the invitation doesn't say so directly. Reading the invitation email carefully for platform names, and doing a quick search on the platform if you don't recognise it, tells you what you're walking into before you start recording, rather than discovering it mid answer.

A step by step routine to prepare for one

1. Confirm the format before you start the clock. Check the invitation for the platform name, whether it's live or recorded, and whether you get one attempt per question or several. Most platforms show this on the landing page before you begin, so read it fully rather than clicking straight through.
2. Test your setup in a genuinely quiet space with good, even lighting facing you, not behind you. Poor lighting and background noise don't just look unprofessional, they can degrade how clearly the system captures your audio, which matters if delivery is part of what's scored.
3. Rehearse two or three likely answers out loud, not just mentally. Use the same structure you'd use in a live interview: a specific situation, what you did, and the outcome. Saying it out loud surfaces filler words and pacing problems that silent rehearsal never catches.
4. Use the practice question if the platform offers one, and actually treat it as practice. Most tools give you an unscored trial question specifically to let you get comfortable with the format, the timer, and hearing yourself on playback. Skipping it to save time is a false economy.
5. Speak to the camera lens, not the screen. It feels artificial at first, but it's the closest equivalent to eye contact this format allows, and it reads as more direct in the recording.
6. Treat the time limit as a structure, not a threat. Most prompts give you enough time for a complete answer if you lead with the point and the specific example, rather than building up to it. If you're consistently running out of time, the fix is usually a tighter opening line, not talking faster.

A worked example

Say the prompt is: "Tell us about a time you managed conflicting priorities." A weak, unrehearsed answer opens with "That's a good question, let me think," then meanders through a vague description with no specific outcome, and runs out of time before landing anywhere. A prepared answer opens immediately with the situation: "In my last role, I had two stakeholders requesting the same sprint capacity for different launches." It states what you actually did: "I built a short comparison of impact and deadline risk for each, and took it to my manager rather than deciding alone." It closes with a specific result: "We shipped the higher risk launch first and the second landed two weeks later with no complaints." That's a complete answer inside most platforms' time limits, and it scores well against both the automated content check and a human reviewer reading the transcript afterward, because it's specific rather than general.

This same discipline, a concrete example over a vague generality, is exactly what makes a resume bullet point work too, not just an interview answer, which is why the two skills reinforce each other; our guide on how recruiters actually use AI to screen your resume covers the equivalent mechanics on the application side, one stage earlier in the same pipeline.

Common questions

What is an AI video interview?

A recorded, one way interview where you answer prompts to a webcam with no live interviewer present, and the recording is scored by software before a human, if one is involved at all, reviews it. It's now faced by 63 percent of job seekers, according to Greenhouse's May 2026 survey.

Does an AI interview replace a real interview with a person?

Usually not entirely. Most employers use it as a first stage filter, with a human reviewing either the full recording or a shortlist the automated pass produces. Whether that human review actually happens quickly, or at all, varies significantly by employer.

How do I know if I'm being evaluated by AI before I start?

Check the invitation for whether it's described as recorded, on demand, or one way, and note the platform name if one is given. Many employers still don't disclose this clearly, so treating any non live, recorded format as likely AI evaluated is a safer assumption than waiting to be told.

What should I do if I don't hear back after completing an AI interview?

Follow up once, directly, after a reasonable window, the same as you would after any application stage. Greenhouse's data found 51 percent of candidates never heard back at all after an AI interview, so a lack of response reflects a common process gap, not necessarily a signal about your specific answers.

Do AI interview tools judge my face and voice, not just what I say?

Some do include delivery and, in fewer systems, facial signals, but the content of your actual answer, specific and directly responsive to the question, remains the largest factor across most current platforms. Rehearsing what you'll say still matters more than how you look doing it.

Is it worth asking an employer if AI is involved in their interview process?

Yes, and it's a reasonable question to ask directly if it isn't disclosed, either in the invitation or when you get a chance to ask the recruiter. It doesn't disadvantage you to ask, and given how many candidates report not being told, asking plainly is a fair and increasingly normal thing to do.

Where ResumeProofed fits, and where it doesn't

ResumeProofed doesn't touch the interview stage at all, live or AI evaluated, and no resume tool honestly can. What it can do is make sure the resume that got you to that recorded interview in the first place was tailored specifically to the role, so the story you tell in your answers lines up with what the employer already read about you, rather than introducing new claims an interviewer, human or automated, has no way to reconcile with your application.

Try ResumeProofed free: two tailored applications on the house, no card required.