AI Is Watching Both Sides of the Job Interview Now
Candidates use AI tools to answer coding questions in real time. Employers use AI to catch them doing it. The people caught in the middle are software engineers trying to find work.
Key points
- AI interview assistants such as Final Round AI and Interview Coder listen to a live technical interview and feed answers to a candidate in real time.
- A Stanford Institute for Human-Centered AI study of 3.4 million real job applications found evidence of adverse impact for Asian and Black applicants when algorithms screened them.
- Meta and AI software platform Factory both allow candidates to use AI openly during technical interviews, grading on reasoning rather than finished output.
- Ginger, an AI voice recruiter built by two software engineers, flags possible AI use by tracking eye movement, response delays, tab switching and speech patterns during early screening calls.
- Archie Payne of technical recruiting firm CalTek Staffing warns that AI detection tools have already flagged strong candidates incorrectly, a false positive that costs companies qualified hires.
Picture a job interview on a video call. The interviewer asks you to write code from scratch. Just off-camera, a second piece of software is already drafting your answer.
That's not hypothetical. Tools like Final Round AI, Interview Coder and ParakeetAI listen through your microphone, process what the interviewer says, and push suggested answers to your screen in seconds. Some claim their overlay is invisible to screen-sharing software.
Software engineers are reaching for these tools most often, and the pressure behind that choice is real. Tech layoffs have piled up. Hiring has slowed. Résumé-screening software, a type of AI that filters out applications before a human reads them, already decides whether many candidates get a first call. Some applicants feel they're fighting an automated process and see AI assistance as evening the odds. Our earlier story on engineers going back to basics to stay sharp captured exactly this anxiety, back in July.
"Companies started to use AI résumé screeners to filter applications at scale," says Archie Payne, president of CalTek Staffing, a technical recruiting firm. "Candidates noticed this and started using AI in their interviews as a countermeasure."
Employers noticed back. Companies now use platforms that watch for signs of AI assistance: unusual pauses, eye movements suggesting off-screen reading, speech patterns that sound generated rather than spontaneous. Ginger, a voice-based AI screening tool built by two software engineers, tracks those signals during first-round calls.
Does AI detection actually work?
Not reliably, not yet. Payne has watched strong candidates get flagged incorrectly. "That can be a serious problem when it can already be a challenge to find people qualified for the position," he says.
The risks go wider than accuracy. A Stanford Institute for Human-Centered AI study followed 3.4 million real applicants, all assessed by algorithms from a single vendor, and found evidence of systematically worse outcomes for Asian and Black applicants. Bias baked into a screening tool scales fast. AI2Day has reported on algorithmic bias three times in the past 90 days, and this study is the sharpest example yet of how screening at scale can quietly entrench discrimination.
Some employers have decided the detection game isn't worth playing. Meta allows AI use during technical interviews. Factory, an AI-native software development platform, goes further. Candidates spend an hour building or migrating real code using AI coding tools, then get graded on how they directed the AI, how they debugged problems and whether they could explain their own solution.
"Weak candidates lean on it to do their thinking and stall the moment it falls short," says Varin Nair, who leads Factory's technical hiring. "Strong candidates use it to move faster and free themselves to reason about architecture, trade-offs, product."
For anyone currently job-hunting: Payne's advice is to use AI to prepare, not to perform. Understand the answers it gives you. Then answer in your own words. Getting caught carries a lasting cost in a field where professional networks are tighter than they look.
The Factory model is the one worth watching. Designing interviews around how engineers actually work today sidesteps the detection arms race entirely, and it probably gives a more honest read on a candidate than a timed coding puzzle ever did.
