I Got Interviewed by an AI Bot for a Job. It Was Worse Than You Think.

A software developer named Chris Schwarz recently sat down for a job interview. No human was on the other end. Instead, he faced an AI-powered bot that asked questions, evaluated his responses in real time, and presumably scored him against some invisible rubric. His detailed account on his blog, SchwarzTech, reads like a dispatch from the uncanny valley of modern hiring — and it should make every tech professional pay attention.

The setup was straightforward enough. Schwarz received a link to a video interview platform. No recruiter, no hiring manager. Just him, his webcam, and an AI agent asking pre-programmed questions about his technical background and experience. The bot listened, followed up, and moved on. Clinical. Efficient. And deeply strange.

Here’s what stood out: the AI interviewer didn’t just recite questions from a script. It appeared to react to his answers, adjusting follow-ups in a way that mimicked conversational awareness. But the mimicry was imperfect. Schwarz described moments where the bot’s responses felt off — slightly misaligned with what he’d actually said, as though it was pattern-matching rather than genuinely understanding. The experience left him questioning whether the system could fairly assess nuance, context, or the kind of soft skills that often determine whether someone actually thrives in a role.

This isn’t a fringe experiment. AI-driven interviewing tools have been gaining traction for years, and the market is accelerating. Companies like HireVue, Paradox, and others have built products that automate early-stage screening through video analysis, natural language processing, and sentiment detection. Wired has reported on how these tools are reshaping hiring pipelines, particularly at large enterprises processing thousands of applicants. The promise: faster screening, reduced bias, lower costs. The reality is more complicated.

Much more complicated.

Schwarz’s account highlights a tension that’s been simmering in HR tech circles. Proponents argue that AI interviews standardize the process — every candidate gets the same questions, the same conditions, the same evaluation criteria. That sounds fair on paper. But critics, including researchers at MIT and NYU, have raised concerns that these systems can encode the very biases they’re supposed to eliminate. An algorithm trained on data from past successful hires will inevitably reflect the preferences — conscious or not — of the humans who made those earlier decisions. MIT Technology Review has covered this problem extensively, noting that bias in AI hiring tools remains one of the most persistent and difficult challenges in applied machine learning.

And then there’s the candidate experience. Schwarz didn’t mince words about how dehumanizing the process felt. Talking to a bot that can’t laugh at a joke, can’t pick up on enthusiasm, can’t read the room. It strips the interview of everything that makes it a two-way evaluation. Candidates aren’t just being assessed — they’re also assessing the company. An AI interviewer sends a message, whether the employer intends it or not: we’d rather automate this interaction than have a human talk to you.

That message lands differently depending on the labor market. In a tight market for software engineers and senior technical talent, it can be a dealbreaker. Schwarz himself seemed put off. Other developers commenting on the post echoed the sentiment — some said they’d withdraw their application on the spot.

So why are companies doing it? Volume. A Fortune 500 company receiving 50,000 applications for a few hundred roles can’t put a human in front of every candidate. The math doesn’t work. AI screening is a triage mechanism, and from a pure operations standpoint, it makes sense. But the tradeoff is real. You gain speed and lose signal. You reduce cost and risk alienating exactly the kind of candidates you most want to attract.

There’s also the regulatory angle. Illinois passed the Artificial Intelligence Video Interview Act back in 2020, requiring companies to disclose when AI is used in video interviews and to obtain candidate consent. New York City’s Local Law 144, which took effect in 2023, mandates bias audits for automated employment decision tools. Reuters has tracked the growing patchwork of regulations, and the EU’s AI Act classifies employment-related AI systems as high-risk, subjecting them to stricter oversight. The legal ground is shifting fast.

What Schwarz’s post captures — better than any policy paper — is the visceral weirdness of the experience. The feeling of performing for a machine that doesn’t care. The uncertainty about what’s actually being measured. The nagging suspicion that the system might be evaluating things it shouldn’t: your background, your accent, the lighting in your room.

These aren’t hypothetical concerns. In 2019, HireVue faced significant backlash over its use of facial analysis in interviews and eventually dropped the feature after pressure from researchers and advocacy groups like the Electronic Privacy Information Center. The company pivoted to audio and text-based analysis, but the episode revealed how quickly these tools can overreach.

For hiring managers and CTOs reading this: the technology works, in a narrow sense. It can filter. It can rank. It can process at scale. But it can’t build rapport. It can’t sell your culture. It can’t tell the difference between a nervous candidate who’d be brilliant on the job and a polished one who’d flame out in three months. Not yet, anyway.

For candidates: if you encounter one of these systems, know that your data is being processed in ways that may not be transparent. Ask questions. Request disclosure. And if the experience feels wrong, trust that instinct — it probably tells you something about the company behind the bot.

Schwarz ended his post with a simple observation. The interview felt like talking to a wall. A very polite, very sophisticated wall. But a wall nonetheless. That’s the state of AI hiring in 2025. Technically impressive. Humanly inadequate.