The AI Interview Trap: When a Robot Decides If You Get the Job

There was a post on Reddit a while back that got something like seventeen thousand upvotes. The person had applied for a job, made it through the initial screen, and then received a link to complete a "video interview." When they clicked the link, there was no interviewer. Just a camera, a set of questions displayed on screen, and an AI system that would analyse their responses — their words, their tone, their facial expressions — and decide whether they'd progress to the next round.

They refused to do it. The comments section overwhelmingly agreed. And i think their instinct was right, but the reality of the job market means the answer isn't always that simple.

AI interviews are becoming increasingly common, particularly for high-volume hiring at large companies. Whether you think they're a good idea (i don't, mostly) is somewhat beside the point. They exist, more companies are using them, and if you're job hunting you're going to encounter them. So let's talk about what they actually are, how they work, and how to think about them strategically.

How AI interviews actually work

There are a few different types and they're worth distinguishing.

Asynchronous video interviews with AI analysis. This is the type from the Reddit post. You record yourself answering predetermined questions. An AI system analyses your responses. The analysis might include natural language processing (what you said), sentiment analysis (how you said it), and in some cases facial expression analysis and body language scoring. HireVue is probably the most well-known platform doing this, though there are dozens of others.

AI chatbot interviews. You interact with a chatbot that asks you questions and evaluates your written responses. These are common for early-stage screening, particularly in customer service and retail roles.

AI-assisted live interviews. A human interviews you, but an AI system is running in the background, scoring your responses in real time and providing the interviewer with suggested follow-up questions. This one's harder to detect because there's still a human in the room.

AI CV screening. This isn't technically an interview, but it's the most common form of AI in hiring. Your CV is parsed by an algorithm that scores it against the job description and determines whether you make it to the interview stage. If you've applied for jobs online and heard nothing back, this is often why.

Why companies use them

The business case is straightforward. A large company might receive thousands of applications for a single role. Screening all of those with human interviewers is expensive and slow. AI tools promise to process them faster, more consistently, and cheaper.

Companies also argue that AI reduces human bias in hiring. There's some theoretical basis for this — humans are definitely biased in interviews — but the evidence that AI actually reduces bias rather than just encoding different biases is, shall we say, not compelling. More on that in a moment.

The honest reason many companies use AI interviews is the same reason they use AI for everything else: it saves money. Whether it produces better hiring outcomes is a question most of them haven't rigorously tested.

The problems (and they're significant)

Bias, but make it algorithmic. AI hiring tools are trained on historical data. If a company's previous successful hires were predominantly white men from elite universities, the AI learns to favour candidates who look like that on paper. Amazon famously had to scrap an AI recruiting tool because it was systematically downgrading CVs that included the word "women's" (as in "women's chess club captain"). The technology has improved since then, but the fundamental problem — bias in training data producing biased outputs — hasn't been solved.

Facial expression analysis is pseudoscience. I'll be blunt about this. The idea that you can reliably assess someone's competence, confidence, or suitability for a role by analysing their facial expressions is not supported by the scientific evidence. It's particularly problematic for neurodivergent candidates, people with facial differences or disabilities, and people from different cultural backgrounds where expressions carry different meanings. Some companies have quietly dropped this feature. Others haven't.

It penalises the wrong things. AI interview systems tend to reward polished delivery, confident eye contact, and fluent speech. These are presentation skills. They're not the same as job competence. A brilliant engineer with social anxiety will score worse than a mediocre engineer with good presentation skills. A native English speaker will score better than an equally qualified non-native speaker. The system is measuring performance in an artificial situation, not ability to do the actual job.

The opacity problem. When a human rejects you after an interview, you can sometimes get feedback. When an AI rejects you, you get nothing. You don't know what the algorithm scored you on, what threshold you didn't meet, or what you could do differently. This makes it impossible to improve and deeply frustrating for candidates.

How to navigate AI interviews (if you choose to)

If you decide to go through with an AI interview, here are the practical tips.

Treat it like a presentation, not a conversation. There's no rapport to build. No back-and-forth. No human to read and adjust to. You're performing for a camera and an algorithm. That means: clear structure, confident delivery, specific examples, and relevant keywords from the job description.

Use the STAR method religiously. Situation, Task, Action, Result. AI systems are typically trained to identify structured responses. Give them what they're looking for. State the situation briefly, describe your task, explain your action in detail, quantify the result.

Mirror the job description language. If the job description says "stakeholder management," use that exact phrase. Not "working with clients" or "managing relationships." The AI is often doing keyword matching, and synonyms don't always register.

Watch your environment. Good lighting, plain background, minimal noise. These seem trivial but they affect how the AI system processes your video. Poor lighting or a busy background can interfere with the facial analysis (if they're using it) and distract the algorithm.

Practice with the platform. Most AI interview platforms let you do a practice recording first. Do it. Watch it back. Adjust. The format is unfamiliar and awkward, and you don't want your first attempt to be the one that counts.

Speak at a moderate pace. Natural language processing works better with clear, evenly-paced speech. Rushing makes it harder for the system to parse your responses. Pausing briefly before answering is fine and probably looks more thoughtful than blurting out an immediate response.

Look at the camera, not the screen. This is counterintuitive but important. If the system is analysing eye contact, looking at the question text on your screen instead of the camera lens will register as avoiding eye contact.

When to refuse

I think there are legitimate reasons to refuse an AI interview.

When it includes facial expression analysis. If you're neurodivergent, have a facial difference, or simply object to pseudoscientific assessment methods, refusing is entirely reasonable. Some jurisdictions are beginning to restrict these technologies specifically.

When you have better options. If you're in a strong position in the job market — good network, multiple opportunities, in-demand skills — you can afford to be selective about which companies you engage with. Refusing an AI interview is a signal that you value human interaction, and some companies will respect that.

When the role is senior. AI interviews are common for graduate and early-career roles. They're much less common for senior positions. If a company is asking a senior candidate to do a one-way AI video interview, that says something about how they value senior talent.

When it doesn't feel right. Your gut reaction to the AI interview process is data too. If a company's hiring process makes you feel like a commodity being processed through a system, that might tell you something about what working there would be like.

If you do refuse, be professional about it. "Thank you for progressing my application. I'd prefer to have a conversation with a member of your team rather than complete an automated video interview. I'd be happy to arrange a call at a time that works." Some companies will accommodate this. Some won't. The ones that don't have told you something about their culture.

The legal picture

AI in hiring is a rapidly evolving legal area.

In the EU, the AI Act classifies AI systems used for recruitment as "high risk," meaning they'll face strict transparency, bias auditing, and human oversight requirements.

In the UK, existing equalities legislation applies to AI hiring decisions. If an AI system discriminates against candidates with protected characteristics, the employer is liable, regardless of whether a human made the decision.

In the US, it varies by state. New York City requires bias audits for automated employment decision tools. Illinois requires consent for AI video analysis in interviews. Other states are likely to follow.

If you believe you've been discriminated against by an AI hiring system, it's worth getting legal advice. The challenge is proving it when you can't see how the algorithm works, but regulators are increasingly pressuring companies to be more transparent about their AI hiring tools.

The future (and it's mixed)

I think AI will remain part of the hiring process. The cost savings are too significant for companies to abandon it entirely. But i also think the worst excesses — facial expression analysis, opaque rejection with no feedback, fully automated decision-making — will gradually be regulated away.

In the meantime, the best thing you can do is understand the system you're operating in and make informed choices about when to play along and when to walk away.

If you've recently been laid off and are navigating a job market full of AI interviews, know that your frustration is shared by millions of people and it's not a reflection of your worth. A system that can't recognise your value isn't measuring the right things. The human on the other side of the process usually knows this, even if the algorithm doesn't.

The one thing to do today: if you're actively job hunting, search for the AI interview platform your target company uses (HireVue, Pymetrics, myInterview, etc.) and watch tutorial videos for that specific platform. Familiarity with the format reduces anxiety and improves your performance. The system might be imperfect, but you can still be prepared for it.