Make talent quality your leading analytic with skills-based hiring solution.
AI assisted cheating in interviews covers two related but distinct threats: candidates using generative AI to feed themselves answers in real time during a live interview, and candidates using synthetic, AI generated video to impersonate someone else entirely. Both have moved from rare edge case to common occurrence extremely fast. Recent research found that 45 percent of candidates now use some form of AI assistance during interviews, and by December 2025, 35 percent had been caught doing so, up sharply from just 15 percent six months earlier.
This is no longer a fringe concern, it is close to the median candidate experience in some hiring pipelines, which makes understanding AI assisted cheating in interviews specifically, not cheating in general, a practical necessity.
The more common version by far is a candidate using a large language model to generate answers during the interview itself. Roughly half of candidates caught cheating used a dedicated tool built specifically for this purpose, and around a third used ChatGPT or a similar assistant through a secondary device, often in voice mode so answers could be read back to them without visibly typing. A much smaller share, about one in five, simply switched tabs or screens to search manually.
This tactic is concentrated in technical roles, where cheating rates reach roughly 50 percent, compared to roughly 10 percent in sales roles, and junior candidates with under five years of experience cheat at close to double the rate of senior professionals.
The rarer but more severe version uses AI generated synthetic video to impersonate a different person entirely, either a more qualified individual or someone whose identity was stolen outright. Roughly 17 percent of HR managers reported encountering deepfake technology in a video interview by mid 2025, and Gartner has projected that as many as one in four candidate profiles could be entirely fabricated by 2028 if current trends continue.
This connects directly to the broader fraud patterns covered in Glider’s existing Deepfake Tech and Candidate Fraud post, but the interview specific version of this threat is what a hiring manager actually encounters live, on a call, in real time.
The data on detection is not encouraging on its own. Sixty one percent of candidates caught using AI assistance still scored above the passing threshold and advanced to the next round before being caught, and a majority of hiring managers believe candidates are currently better at faking than recruiters are at catching it.
This is consistent with what our companion piece on AI based test monitoring found for written assessments, single detection methods used in isolation are not reliable on their own.
Verify identity at the start of the interview using real identity verification and liveness detection, since this specifically defeats the deepfake half of this threat. For the chatbot half, ask unscripted, highly specific follow up questions that a real time prompting tool cannot anticipate well, and watch for the fluency and pause patterns described in our companion interview cheating guide.
And where the role justifies it, run interviews through a structured, monitored live interview platform rather than an unmonitored video call, since both threats described here specifically exploit the lack of any verification layer in a standard call.
AI assisted cheating in interviews is not a single problem with a single fix, it is two distinct threats, a chatbot feeding answers and a deepfake feeding a false identity, that happen to converge on the same moment: the live interview. Naming them separately, and defending against each specifically, is more effective than treating AI cheating as one vague, unsolvable category.
AI assisted cheating in interviews covers two distinct threats: candidates using AI chatbots to generate answers in real time during a live interview, and candidates using AI generated synthetic video to impersonate a different person entirely.
Recent research found that 45 percent of candidates now use some form of AI assistance during interviews, and 35 percent had been caught doing so by December 2025, up from 15 percent just six months earlier.
About half of caught candidates used a dedicated real time prompting tool built for this purpose, about a third used ChatGPT or a similar assistant through a secondary device, and a smaller share simply switched browser tabs to search manually.
Deepfakes can be detected using liveness detection technology that analyzes texture, lighting, and involuntary movement patterns synthetic video lacks, though detection has not fully kept pace with the sophistication of newer deepfake generation tools.
Employers can verify identity and use liveness detection at the start of the interview to address deepfake impersonation, and ask unscripted, specific follow up questions that real time prompting tools struggle to anticipate, to address chatbot assisted cheating.
Assessment cheating covers the ways a candidate can produce a test result that does not reflect their own, unaided ability on a written or coding assessment. Unlike interview cheating, which happens in a live conversation, assessment cheating happens against a clock and a fixed question set, which shapes a different set of tactics and a […]
Interview cheating covers the tactics a candidate uses to misrepresent who they are or what they know during a live conversation, as opposed to a written test. Remote video interviews made this meaningfully easier, since a screen removes several of the physical cues that used to make cheating obvious in person. Understanding the specific tactics […]
Candidate Identity Checks: Why They Are No Longer Optional Candidate identity checks used to be treated as an extra step reserved for high security roles, government contracts, financial services, healthcare, while most employers relied on a resume, an interview, and a standard background check. That assumption is breaking down quickly, for reasons that have nothing […]
