Picture a job interview. Now picture that the person conducting it is an animated avatar. Now picture that the candidate answering its questions might be one too. That’s not a hypothetical anymore — it’s just what hiring looks like in a growing number of companies right now, and both sides are increasingly uneasy about it.
AI-driven hiring has quietly gone from a novelty to a default. According to SHRM’s 2026 State of the Workplace survey, 43% of HR leaders now use some form of AI in their interview process, up from just 17% in 2023 — a shift from experimental to standard practice in roughly three years. Some of that AI shows up as an animated avatar conducting the interview itself, asking questions and scoring responses with no human present at all. And in a twist that says a lot about where this technology is headed, some of the candidates on the other side of that conversation are turning out to be AI avatars too.
Why Companies Started Doing This
The pitch to employers is straightforward: AI-led interviews let a company screen every applicant instead of the small subset a human recruiting team could realistically get through. Platforms like CodeSignal, Humanly, and Eightfold have built entire products around this promise — natural language processing handles the conversation, speech recognition transcribes it in real time, and machine learning models score responses against patterns pulled from a company’s previous successful hires. Some systems go further, using computer vision to analyze body language and facial expressions during the interview itself.
That last piece is where things start to get genuinely contentious, and for good reason. Reading tone of voice and word choice is one thing; inferring personality or trustworthiness from facial expressions and posture is a different, far shakier proposition — and the research backs up the skepticism. Recent industry analysis places the predictive validity of full personality inference from interview video under 0.20, a figure worse than what a well-run, structured human interview typically achieves. In plain terms: the AI reading your face during an interview may be doing worse than a competent human would, while looking a great deal more scientific.
The Bias Problem Nobody Fully Solved
AI-led hiring was sold, in part, on the promise that it would be less biased than a human interviewer. The evidence on that front is genuinely mixed. A 2026 study published at the CHI Conference on Human Factors in Computing Systems found that people react unconsciously to an AI avatar’s appearance even when they know full well they’re talking to a machine — and that this reaction intensifies specifically at the moment of rejection. Test subjects who received a rejection and were later interviewed again by an avatar were more likely to believe they hadn’t been judged fairly, particularly when the avatar shared just one visible trait with them. The researchers’ framing is worth repeating directly: we all unconsciously react to an avatar’s appearance, even knowing it isn’t a person.
That’s a genuinely uncomfortable finding for an industry that leaned on “AI is less biased” as its core sales pitch. Swapping a human interviewer for an avatar doesn’t remove the psychology of bias from the room — it just changes what triggers it.
Companies themselves seem to sense this tension. A recent industry survey found that while a third of companies expect AI to effectively run their hiring process by 2026, 71% still retain some form of human oversight, and more than three-quarters have implemented policies specifically governing the ethical use of AI in hiring, with another fifth planning to. That’s not the behavior of an industry that’s fully confident in the technology it’s deploying — it’s the behavior of one hedging against a tool it isn’t sure it can fully trust yet.
Then Candidates Started Fighting Back With the Same Tools
Here’s where the story stops being a simple “AI vs. human fairness” debate and turns into something stranger: candidates have started showing up to interviews as AI avatars of their own.
Recruiters are already describing the experience. Sydney-based recruiter Lana Kersanava told Semafor that these digital renders — mirroring a real applicant’s voice and likeness while being, underneath, entirely artificial — have shown up in some of her interviews over the past six months. “At the beginning, it’s like you feel like something is off,” she said. “It looks very realistic. But it’s not.”
The scale of this isn’t small anymore. A 2026 Greenhouse report found that 91% of US hiring managers have encountered or suspected AI-generated answers during online interviews, and across nearly 19,400 live interviews analyzed between mid-2025 and early 2026, 38.5% of candidates were flagged for AI-assisted behavior. Security researchers at Palo Alto Networks found it takes as little as 70 minutes for someone with zero image-manipulation experience to build a convincing fake candidate capable of passing a video interview. And it isn’t just amateur fraud — real-time face and voice cloning that used to require Hollywood-level production has become achievable with off-the-shelf tools, to the point that a demonstration by security firm Pindrop transformed a reporter’s face live on a video call and generated a voice clone capable of holding an unscripted conversation.
By 2026, these fabricated interview appearances have gotten alarmingly convincing — blinking, smiling, and nodding at the right moments, with only subtle glitches like lip-sync delay or slightly unnatural eye movement giving the game away. Gartner projects that by 2028, one in four job candidate profiles globally will be fake in some form, spanning AI-generated profiles, deepfake video interviews, and fabricated work histories entirely.
The response from some major employers has been almost old-fashioned: the Wall Street Journal has reported that companies including Google and McKinsey began reintroducing mandatory in-person interviews specifically to counter the surge in AI-assisted interview fraud. When the technology arms race gets sophisticated enough, the most reliable countermeasure turns out to be a conference room and a physical handshake.
What’s Actually Worth Trusting Right Now
Not every AI hiring tool deserves the same level of suspicion, and lumping them all together misses an important distinction. Industry analysis breaks the current landscape into a few genuinely different categories, and they don’t carry equal risk. Scheduling and coordination assistants and async video platforms that simply organize logistics are relatively uncontroversial. Fully autonomous, no-human-present “interview avatars” — the kind actually conducting entire interviews start to finish — carry both poor candidate-experience scores and real legal exposure, particularly under the EU AI Act’s high-risk classification for automated decision systems. Their practical niche, for now, seems to be extreme-volume hourly hiring rather than any role requiring real judgment. Full personality inference from video is the shakiest category of all, given how thin the predictive science actually is.
The picture that emerges isn’t “AI hiring good” or “AI hiring bad” — it’s that some categories of these tools are earning their place in a legitimate hiring process, while others are still mostly marketing dressed up as science, and a few create genuine legal and fairness risk if a company deploys them without real guardrails.
Where This Leaves Everyone
There’s a phrase making the rounds among hiring platform executives that captures the moment well: hiring has become stuck in what Greenhouse’s CEO has called an “AI doom loop” — employers deploying AI to handle a flood of applications, partly created by candidates using AI to apply to more jobs faster, partly by candidates now also using AI to interview convincingly for those jobs, which pushes employers toward even more AI-driven screening to keep up. Each side’s adoption of the technology accelerates the other’s.Nobody involved seems entirely comfortable with where this is heading, and the data backs that discomfort up: rising fraud detection rates, active regulatory hardening in the EU and several US states, and major employers quietly walking back toward in-person interviews as a fraud countermeasure all point the same direction. The tools aren’t going away. But the confident, “this is obviously the future” phase of AI hiring adoption looks to be giving way to something more cautious — a period where companies are learning, sometimes the hard way, exactly which parts of this technology they can actually trust.
