How to Succeed at a Video Interview (Live and AI Async)

A video round used to mean one thing: a webcam call with a recruiter. That is still common. But it is no longer the only thing.

A growing share of Dutch graduate programmes (banking, consulting, Big 4, large corporates) now ask you to record yourself answering pre-set questions, alone, on a platform like HireVue. There is no live interviewer. The video sits there until a recruiter, or a scoring system, gets to it.

Both rounds are called video interviews. They feel completely different in practice. Most generic advice covers only the live one, which leaves a lot of graduates walking into the recorded version with no real plan.

This guide covers both. Live first, because it is still the most common first-round screen for NL employers. Then async AI, with the parts that actually matter: how the platforms work, what is and is not being scored, and what the EU AI Act now requires.

TL;DR

There are two video formats you might face: live video (Teams, Zoom, or Meet with a real recruiter on the other end) and async AI video (you record answers alone, usually on HireVue or a similar platform). Most graduates will see one or the other within a single application process.

For live: webcam at eye level, light from the front, structured short answers, look at the lens when it matters. Most mistakes are setup, not content.

For async AI: the typical structure is three to five questions, with thirty seconds to prepare and two to three minutes to answer each one. Verbal content carries the score. Visual scoring of facial expressions is no longer standard, and the EU AI Act treats hiring AI as high-risk from 2 August 2026.

STAR format works in both. In async, it works extra well because clear structure is easier to score, by humans and AI.

There are two distinct video rounds, and they ask different things from you.

Live video

A scheduled call with one or more interviewers, on Microsoft Teams, Zoom, or Google Meet. This is still the most common first-round screen for graduate roles in the Netherlands. Generally a recruiter or junior HR contact, 25 to 45 minutes, broadly the same as a phone screen but with cameras on.

You can ask clarifying questions, you get reactions in real time, and tone or rapport is part of how you come across.

Async AI video (also called one-way video)

You log into a platform (HireVue, AsyncInterview.ai, or similar), see a question, get a short prep window, then record your answer to camera. No interviewer is on the line. Your responses are reviewed later by a recruiter, often supported by an AI scoring layer.

Typical structure: 3 to 5 questions, around 30 seconds preparation, up to 2 or 3 minutes per answer. Sometimes you get one retake per question, sometimes none. Most platforms give you a 5 to 7 day window to complete the full set.

How to tell from the invite

Live: the email contains a calendar invite, a meeting link, and the names of one or more interviewers.

Async: the email contains a link to a platform (HireVue is the most common, often with .hirevue in the URL), a deadline ("complete within 5 days"), and no specific meeting time. Sometimes the wording is "on-demand interview" or "video assessment."

The seven differences that actually matter

Where each format shows up. Live video is the default for most NL first-round screens, used by SMEs and large employers alike. Async AI video is more common in graduate programmes at scale, particularly banking, Big 4 traineeships, large corporates, and consulting. Think Rabobank, ABN AMRO, JPMorgan, Diageo, BCG, and Deloitte.

The format itself. Live is a scheduled call on Teams, Zoom, or Meet, running twenty-five to forty-five minutes. Async is a series of recorded answers on a platform like HireVue, with no interviewer present.

Prep window. Live gives you as long as you want before the call. Async gives you thirty seconds (sometimes sixty) per question, after the question appears on screen. That window is fixed by the platform, not the employer, so it cannot be negotiated.

Answer length. Live answers are conversational, most run sixty to one hundred and twenty seconds. Async answers are capped, usually at two to three minutes, and you should use the full time. HireVue defaults to three minutes per question.

What is being assessed. Live is judging communication, fit, motivation, judgment, and rapport. Async is judging verbal content and structure mainly, looking at language patterns and keywords. Async scoring is human-led, often AI-assisted.

Recovery if you stumble. On live, pause, restart the sentence, and move on. Easy. On async, recovery depends on retake rules, some platforms allow one retake per question, some allow none. Always check the retake rules before you start the first question.

Interaction. Live is two-way: you can ask the recruiter to repeat or clarify. Async is one-way: question, prep, record, no follow-ups. Async cannot be rescued with charm, so structure has to carry the answer.

This is not a hierarchy. Neither format is harder by definition. They reward different preparation.

The recruiter is not scoring your bookshelf. They are scoring whether they can see and hear you, and whether you look like you took the conversation seriously.

Camera

Position the webcam at eye level. If you are on a laptop, that usually means putting it on a stack of books or a small box. A laptop on a desk shoots up your nose, which is a bad look on anyone.

Frame yourself from mid-chest up. Your eyes should sit roughly one third from the top of the frame. Leave a small amount of headroom, not too much.

Light

Light should come from in front of you, not behind. Sitting with a window at your back turns you into a silhouette. A window facing you, or a desk lamp pointed at the ceiling and bouncing back, works fine.

Avoid mixed colour temperatures (one warm bulb plus one cool bulb pointed at the same spot). It looks strange on camera and the auto white balance flickers.

Audio

Audio matters more than video quality. A recruiter can forgive a soft-focus webcam, but they cannot forgive not understanding you.

Wired earbuds with a mic beat your laptop microphone almost every time. They isolate your voice from room echo. Bluetooth earbuds work but can drop, especially during long calls. Test both before you commit.

Connection

Wired ethernet is the safest option if you have a port. Wi-Fi is fine in most homes, but sit close to the router and ask anyone at home not to start a 4K stream during your call.

Have your phone next to you with a hotspot ready, in case home Wi-Fi fails. Do not panic if it drops. Reconnect, apologise once, move on. Recruiters know it happens.

Background

A plain wall is best. A tidy bookshelf is fine. A virtual blur is fine when the wall behind you is not presentable. Avoid novelty backgrounds (beach, space, anything moving). They distract.

What to wear

Smart-casual is the Dutch graduate norm. A clean shirt or blouse, sometimes a blazer for finance or law roles, almost never a tie. Match the level of the interviewer if you can tell from LinkedIn.

Solid colours work better than busy patterns on camera. Patterns can shimmer or moiré on lower-quality streams.

Live video makes everyone look a little less expressive than they really are. The fix is small adjustments, not theatre.

The eye-contact problem

On a video call, your eyes can be in two places: the recruiter's face on the screen, or the camera lens. They cannot be in both. If you look at the screen, you see their reactions but appear to look down. If you look at the lens, you appear engaged but cannot see them.

A 2025 study in the Journal of Business and Psychology found that horizontal deviation of gaze (looking off to one side) hurt hirability ratings, while a small vertical deviation (looking just below the lens) did not. The takeaway: vertical glances at the screen are forgivable. Horizontal ones look distracted.

The simple fix: shrink the meeting window and drag it to the top of your screen, directly under your webcam. Now when you look at the recruiter's face, you are also looking close to the lens.

Energy

Webcams flatten things. Your normal calibration of gestures and tone reads as slightly muted on a recruiter's screen. Speak with a touch more clarity than you would in person. Keep gestures inside the frame.

Smile when you mean it. A polite resting expression is fine. A constant fixed smile is not. Recruiters notice.

Pacing

Most people speak faster on video than they realise. Add a small breath between sentences. It feels long to you, normal to the listener.

If you need a moment to think, say so. "Let me think for a second" is a complete sentence. Two seconds of thinking sounds professional. Three seconds of dead air sounds like a connection drop.

Structure

Use the STAR format for behavioural questions. Situation in one or two sentences, Task in one, Action in three or four, Result in one or two. Sander's deeper guide on STAR (and the Dutch STARR variant) covers the mechanics: [LINK PLACEHOLDER: post #12 STAR for Behavioural Questions].

On video, structure matters more than in person, not less. The interviewer is processing your answer through a small window of your face. Anything that helps them follow the logic is gold.

Async video is not a fad. According to SHRM's State of AI in HR 2026 report, recruiting is the largest single use case for HR-related AI, and AI-supported video assessment is one of the fastest-growing tools inside that. In Dutch graduate programmes, async video has been standard for several years at firms running large applicant volumes.

Who uses it in the Netherlands

Banking traineeships, including JPMorgan, Goldman Sachs, and (often via Pymetrics-style game assessments paired with HireVue) ABN AMRO and Rabobank. Big 4 audit and consulting graduate intakes (Deloitte, EY, KPMG, PwC). Large multinationals with NL offices like Diageo, Microsoft, and Amazon. Strategy consulting firms (Bain, BCG, McKinsey use a mix of async and live).

If you are applying to a smaller Dutch employer or a startup, you almost certainly will not see one. They are too expensive to deploy at low volumes.

What the format actually is

You log in. The platform shows you a question (often as text, sometimes as a pre-recorded video of someone reading it). A countdown starts: typically 30 seconds to think. Then a record button appears, you have a fixed time to answer (usually 2 to 3 minutes), and the platform stops you when time runs out.

Some platforms allow one retake per question. Some allow none. The retake policy is shown before each question. Read it. Decide your strategy before you click record on the first one.

What gets scored, and what does not

Modern AI video tools score verbal content and language patterns: the words you use, the structure of your answer, keywords matched against the role, response length, how directly you address the question.

HireVue removed facial-expression analysis from default configurations in 2021 after regulatory and academic pressure. Several platforms have followed. As of 2026, the EU AI Act explicitly bans emotion recognition in workplace and education settings (Article 5), with that prohibition already in force from February 2025. So if a vendor still pitches "micro-expression analysis" for hiring, that is a vendor with a legal problem, not a feature you need to defeat.

The remaining AI scoring is mostly pattern matching on language. It rewards clear structure, on-topic answers, and competency keywords drawn from the job description. It does not reward big personality, jokes, or charm.

Practice protocol

Record yourself, on your laptop, answering generic versions of the questions. Give yourself the same 30-second prep window and the same 2 to 3 minute cap. Watch the playback.

Most people are shocked by their own footage. Pace, pauses, fillers, that one expression you make when you are thinking. Three runs through your own playback is worth ten hours of reading interview tips.

Start with three or four common graduate prompts: "Tell us about yourself", "Tell us about a time you led a team", "Why this company", "Describe a challenge you faced".

STAR works extra well here

On a live call you can read the room and adjust. On async, you cannot. Structure is the only thing keeping a 2-minute answer from rambling. STAR is the structure that nearly every scoring rubric, human or AI, is calibrated to.

Situation: 1 to 2 sentences. Task: 1 sentence. Action: 3 to 4 sentences, the bulk of the answer. Result: 1 to 2 sentences with a number where you can. Practice fitting one full STAR answer in 90 to 120 seconds. That timing is your sweet spot for a 3-minute cap.

The talk-to-the-lens technique

Tape a small sticky note next to your webcam with a smiley face on it. Speak to the smiley face. It sounds silly. It works. Without a human on the other end, your face naturally drifts toward the screen, looking for one. Giving it a target keeps your gaze on the lens, which is what the recording sees.

Common pitfalls

• Rambling past the time limit. The platform cuts you off. If your conclusion was at the end, the recruiter never hears it. Practice landing the result inside the cap.
• Over-rehearsing. Memorised answers read flat on camera. Know the structure, not the script.
• Treating it like a casual conversation. Without a human, candidates sometimes drop into a too-relaxed register. Slightly more formal than a friend chat, slightly less formal than a written essay.
• Skipping retakes when allowed. If the platform offers one retake and your first take had a stumble in the first 10 seconds, retake it. The retake is not a flag. It is part of the format.

Async AI video has been criticised, including by serious academic and regulatory bodies, for reasons that are worth understanding. Glossing over them would be unhelpful.

What the concerns actually are

AI scoring models are typically trained on data from previous hires. If those previous hires skewed in a particular direction (gender, ethnic background, accent, education path), the model can quietly learn to prefer the same pattern. This is the well-documented bias-replication risk in any historical training set.

Vendor transparency is uneven. Some platforms publish detailed methodology and bias audits, others do not. Without that, candidates and even employers cannot fully verify what is being weighted.

There is also a candidate-experience concern. Recording into the void, with no feedback, can feel cold. That hits hardest for graduates already anxious about a first job market.

What protects you (in the EU)

More than in most parts of the world. Recruitment AI is classified as high-risk under Annex III of the EU AI Act. From 2 August 2026, deployers (employers using the tools) have a documented set of obligations: bias testing on training data, ongoing output monitoring, human oversight on consequential decisions, transparency to candidates that AI is being used, and logged records of AI-assisted decisions.

Penalties scale up to 15 million euros or 3 percent of global turnover for high-risk system non-compliance. That is the kind of number that focuses procurement teams. Major vendors are racing to publish compliance documentation; ones that cannot will have a hard time selling to EU employers.

On top of that, GDPR Article 22 already gives you the right not to be subject to a fully automated decision with significant effects. In practice, no responsible NL employer rejects on AI alone. There is always a human in the loop.

What you can actually do

• Ask. Recruiters expect questions like "Will my recording be reviewed by AI as well as a person?" or "How long is the video stored?". Polite, normal, increasingly required to answer.
• Optimise for what is actually scored. Verbal content and structure carry the score. STAR, role-relevant keywords drawn from the job description, clear answers to the question asked. That is in your control.
• Use accommodations if you need them. If a disability or language situation makes async video harder for you, ask the recruiter for an alternative. "Could I do a live call instead?" is a legitimate ask, and EU AI Act human-oversight rules support it.
• Do not catastrophise. The system is not a black box analysing your soul. It is a structured scoring rubric you can prepare for, with a human reviewing the output.

After a live round

Send a short follow-up email within 24 hours. Two to three sentences. Thank the interviewer, mention one specific thing that came up that you found interesting, restate your interest. Do not write a fresh cover letter.

Then wait. Most NL graduate processes give a decision within 5 to 10 business days. If you have heard nothing after two weeks, a polite check-in is fine.

After an async round

Silence is normal and longer. Recruiters batch async reviews, often after the recording deadline closes for everyone. You may not hear anything for 2 to 4 weeks.

There is usually no follow-up email expected. There is no live recipient, so a thank-you reads oddly. Save your follow-up energy for the next round, when there is a human to thank.

Where this fits in the broader process

In Dutch graduate programmes, the video round is rarely the only stage. It is one piece of a process that often includes a CV/cover-letter screen, sometimes a cognitive or aptitude test, then later a hiring manager or assessment-day round. The orientation guide for that whole process: [LINK PLACEHOLDER: post #10 How to Prepare for a Job Interview in the Netherlands].

If your video round was paired with an aptitude test, the dedicated guide on those covers what providers (SHL, Cubiks, cut-e) are looking for and how to practice efficiently: [LINK PLACEHOLDER: post #13 Cognitive or Aptitude Test].

A video round, in either format, rewards practice you can do on your own laptop. The hard part is making yourself do it.

Aurora, GradGuide's free AI career coach, is built to be your practice partner. You can run a mock recruiter call, generate likely behavioural questions for the specific role you applied to, simulate an async AI question and record yourself against it, and draft your 24-hour follow-up email. Three or four laps with her before the real round, and the real round feels like a fifth take.