What Is an AI Interview? How It Works & the Cheating Crisis

Last Updated: September 28, 2026

Summary: An AI interview is a job interview conducted or scored by AI, either through a one-way video interview a candidate records alone, or a live AI interviewer that asks questions and evaluates responses in real time, using platforms like HireVue. The twist most guides miss: a 2026 study of over 19,000 interviews found 38.5% of candidates showed signs of using AI to cheat during the interview itself, and 61% of them still scored high enough to advance.

AI has become infrastructure on both sides of the hiring table. HireVue's 2026 Global AI in Hiring Report, based on a survey of over 3,100 hiring managers, found 77% of HR teams now use AI regularly in hiring, while only 41% fully trust its decisions. This guide covers how AI interviews actually work, on both the employer and candidate side, and the interview-cheating problem that's grown fast enough to change how companies run interviews at all.

💡 Not sure which AI tool is actually right for your business?

Get the free guide, Which AI Tool Should You Actually Use? — a straightforward breakdown of the leading AI tools to help you pick the right one for your needs.

Subscribe to AI Business Weekly for the guide, plus daily coverage of AI trends, acquisitions, and product launches.

Get the Free Guide →

An AI interview covers two related but distinct things, and the term gets used for both without much distinction, which is part of why the category is confusing. On the employer side, an AI interview is a job interview where AI asks the questions, evaluates the candidate's answers, or scores a recorded response, sometimes with no human involved until a shortlist is produced. On the candidate side, an "AI interview assistant" or "AI interview copilot" is a tool a candidate uses to help them prepare for, or in some cases secretly get help during, a human-conducted interview.

Both meanings are growing fast, and they're connected: as more employers automate parts of the interview process, more candidates have turned to AI tools to compete with automated screening or with each other, which is exactly where the more contentious side of this category comes from.

Most employer-side AI interview tools fall into two formats. A one-way video interview has the candidate record answers to preset questions alone, on their own time, and an AI model then scores the responses against traits the employer is screening for, sometimes analyzing word choice, tone, and specific answer content, before a human ever reviews the recording. A live AI interviewer, a newer and less common format, actually conducts a real-time conversation with the candidate, asking follow-up questions based on what the candidate says, closer to a live interview than a recorded assessment.

Platforms like HireVue built the category around the one-way video format specifically, and the practical case employers make is consistency and speed at scale, every candidate answers the identical questions under identical conditions, and a recruiting team can screen far more candidates than a live-interview-only process would allow. The tradeoff is the one AI hiring critics raise consistently: an AI system trained on past hiring patterns can encode and repeat historical bias, which is a real, documented risk worth understanding on its own. For a deeper look at that specific concern, see our guide on AI hiring discrimination.

Where a company sits between these two formats generally tracks hiring volume. A one-way video interview makes the most economic sense for high-volume roles, retail, call center, hourly, and entry-level hiring, where hundreds or thousands of applicants apply for a smaller number of openings and a human recruiter watching every recorded answer isn't realistic. A live AI interviewer or a fully human interview process remains far more common for senior, specialized, or leadership roles, where the number of candidates is small enough that speed matters less than depth, and the cost of a bad hire is high enough that most companies still want a person making the final call.

This is the part almost no "AI interview" guide actually covers, and it's arguably the bigger story in this category right now. Fabric's analysis of 19,368 real interviews conducted between July 2025 and January 2026 found that 38.5% of candidates showed measurable signs of using AI to cheat during the interview itself, reading or being fed AI-generated answers in real time rather than answering on their own.

The breakdown is more specific than a single headline number. Technical roles saw a 48% cheating rate, four times higher than sales roles at 12%. Junior candidates with 0-5 years of experience cheated at nearly double the rate of senior candidates, and the study found a clear negative correlation between compensation level and cheating rate. Among candidates interviewed more than once, 30% cheated in every interview, 47% never cheated at all, and the rest cheated situationally. The most common method, used in 45% of detected cases, was a dedicated cheating assistant tool built specifically for this (Cluely and Interview Coder are named examples), followed by voice-mode general AI assistants like ChatGPT at 34%, with old-fashioned tab-switching to search answers accounting for 18%.

The part that should concern any hiring manager reading this: 61% of candidates flagged for cheating still scored above the interview's pass threshold, meaning most undetected AI-assisted answers were good enough to advance the candidate anyway. Cheating rates also weren't stable, they roughly tripled between July and September 2025 alone, suggesting this is a fast-moving problem rather than a settled one.

The best-known dedicated tool in this category is a real example of how fast the space has moved and how openly it started. Cluely, founded in April 2025 by Roy Lee and Neel Shanmugam shortly after their suspension from Columbia University over a similar earlier tool called Interview Coder, launched with the explicit tagline "Cheat on Everything" and marketing that named job interviews directly as a use case. By late April 2025, facing public scrutiny, the company scrubbed the cheating-focused language from its site, and by November 2025 it had repositioned as a general AI meeting assistant comparable to Otter.ai. The product's underlying capability, an invisible overlay that reads screen and audio content and generates real-time answer suggestions, hasn't fundamentally changed, only the marketing around it.

The response from employers has split into two camps, and neither is a clean solution yet. Some companies have moved toward requiring on-camera, proctored interviews specifically to prevent a candidate from reading off a second screen, treating it closer to how a proctored exam works. Others have leaned further into AI themselves, using detection tools that flag unnatural pauses, eye movement toward a second monitor, or answer patterns that look AI-generated rather than personally recalled.

Neither approach fully solves the problem the research points to. Proctoring catches obvious tab-switching but not a voice-mode AI assistant a candidate is listening to through an earpiece. Detection tools flag suspicious patterns but, per the 61% pass-through rate above, plenty of AI-assisted answers are good enough to look like a strong, unassisted response. The practical shift several companies have made instead is weighting in-person or live technical assessments more heavily for roles where the 48% technical-role cheating rate makes remote screening least reliable, treating the initial AI-scored or remote round as a coarser first filter rather than the final word.

The employer side of this category is priced for companies, not individuals, and the range is wide enough to matter for anyone evaluating it. A pricing breakdown of HireVue's plans shows an Essentials tier running $25,000-40,000 a year, a Standard tier at $40,000-100,000, and Enterprise pricing from $100,000-145,000 or more, with a specific catch worth knowing: AI-based candidate scoring, the feature most teams actually buy the platform for, only unlocks at the Enterprise tier. HireVue also isn't a standalone applicant tracking system, it plugs into an existing ATS like Greenhouse, Workday, or iCIMS, so the cost is additional on top of whatever a company already pays for its core hiring software.

That price point explains part of why AI-conducted interviews concentrate at large employers and high-volume hiring rather than small businesses. A company running a handful of hires a year has little reason to pay six figures for AI screening capacity; the platforms make the most economic sense for organizations processing hundreds or thousands of applications where consistent, fast first-round screening genuinely saves recruiter hours at scale.

Not every AI interview tool is built around undetected cheating, and conflating the two understates a real, legitimate category. AI interview assistant tools built for practice, generating realistic mock questions for a specific role, giving feedback on answer structure, or helping a candidate rehearse out loud before a real interview, serve a genuinely different purpose than a live-answer-feeding copilot, even though search terms for both overlap heavily.

The distinction that actually matters is timing and disclosure. Using an AI tool to practice answering common interview questions the night before, the same way a person might rehearse with a friend, is a different thing than an undisclosed AI feeding real-time answers during the actual interview. Most hiring managers have no objection to the first category and a clear objection to the second, and the research above suggests the second category is the one growing fast enough to change how interviews get run.

The practical value of legitimate prep tools comes from repetition a person can't easily get otherwise. A human can rehearse with a friend or mentor a handful of times before a real interview; an AI practice tool can run through dozens of variations of the same question, flag filler words or a rambling answer structure, and let a candidate iterate privately without burning a favor from someone in their network. That's a genuinely different value proposition than a live-answer copilot, even though both get marketed under overlapping "AI interview assistant" language, which is part of why the category is hard for a candidate or employer to evaluate at a glance without knowing which specific tool does what.

Using AI to prepare beforehand, practicing answers, researching the company, is not generally considered cheating and is functionally similar to any other interview prep. Using AI in real time during the interview to generate or feed answers without disclosure is what a 2026 study found in 38.5% of over 19,000 real interviews, and it's the practice most companies are actively trying to detect and prevent. The distinction is timing and disclosure, not whether AI was involved at all.

A study analyzing 19,368 real interviews found 38.5% of candidates showed measurable signs of AI-assisted cheating, with technical roles at 48%, four times the rate seen in sales roles. The rate roughly tripled between July and September 2025, suggesting it's a fast-growing problem rather than a stable one. About 61% of detected cheaters still scored high enough to advance in the hiring process.

Common detection methods include on-camera proctoring to catch a candidate reading from a second screen, and AI-based detection tools that flag unnatural pauses, eye movement patterns, or answer phrasing that looks AI-generated. Neither method is fully reliable. Voice-mode AI assistants a candidate can listen to without visibly reading anything are harder to detect than a candidate visibly switching browser tabs.

A traditional one-way video interview has a candidate record answers to preset questions, which are then scored by AI after the fact, often before a human reviews anything. A live AI interviewer actually conducts a real-time conversation, asking follow-up questions based on what the candidate says, which is a newer and less common format than the pre-recorded video style most companies use today.

Pricing runs into enterprise territory. A breakdown of HireVue's pricing shows plans from $25,000 a year at the entry level up to $145,000 or more at the Enterprise tier, and the AI-based candidate scoring most teams actually want typically only unlocks at that top tier. HireVue also requires an existing applicant tracking system to plug into, so it's an additional cost on top of core hiring software, which is why this category concentrates at large employers running high-volume hiring rather than small businesses.

It's a real, documented concern separate from the cheating issue. An AI system trained on historical hiring data can encode and repeat past bias in who gets advanced, which is a distinct risk from interview cheating and worth understanding on its own terms. See our full guide on AI hiring discrimination for a closer look at that specific risk.

An AI interview now describes both an employer screening tool and a candidate-side prep or cheating tool, and both sides of the category are growing fast. The part most coverage misses is the scale of the cheating problem specifically: nearly 4 in 10 candidates in a real 19,000-interview study showed signs of AI-assisted cheating, and most of them weren't caught before advancing. Whichever side of the interview table someone is on, understanding that gap, not just the convenience AI adds, is the part worth actually knowing before the next interview happens.

By Sameer Khan

This article was AI-assisted, then reviewed by Sameer Khan before publishing.

Sameer Khan is the founder of AI Business Weekly. He has a background in research and advisory, working with HR leaders and executives across Canadian public-sector and enterprise organizations on research and AI adoption. He holds an MBA from the Ted Rogers School of Management and has spent nearly a decade in B2B sales across SaaS, research and advisory, and AI.