The New Interview Cheating Economy: What Hiring Teams Are Actually Up Against in 2026

Three years ago, cheating in a remote interview meant glancing at notes off-camera or quietly opening a second tab. The interviewer's instincts — watch the eyes, ask a follow-up — were usually enough. That era is over.

What replaced it is something most hiring teams haven't fully registered: a funded, fast-moving consumer software industry built specifically to feed candidates answers during live interviews, engineered from the ground up to be invisible to the interviewer. This isn't a fringe activity. It's a category with venture backing, ten-million-user products, and a pricing model that makes it rational for almost any candidate to try.

If you hire remotely, this is now part of your risk surface whether you've accounted for it or not. Here's what's actually happening.

Cheating became a product, then an industry

The clearest signal of how far this has come is where the money went. Cluely — a tool that began as "Interview Coder," built by a student who was suspended from Columbia after posting himself using it in an Amazon interview — raised a $15M Series A led by Andreessen Horowitz in mid-2025, on top of a $5.3M seed. When a top-tier venture firm funds the cheating side of the market at that scale, the category signal is unambiguous: this is a business, not a prank.

And Cluely is one of dozens. Final Round AI reports more than ten million users. A continuous stream of competitors — LockedIn AI, ParakeetAI, Leetcode Wizard, Beyz, Sensei AI, Verve AI, and many more — has launched in the past eighteen months, most priced between $20 and $60 a month. Against a six-figure target salary, that price is trivial. The economics guarantee this is a permanent category, not a passing trend, and every new entrant expands the surface a hiring team has to account for.

Underneath the brand names, these tools are strikingly similar. They capture the interview — the spoken question, the coding prompt on screen, sometimes the candidate's own resume for personalization — send it to an AI model, and return an answer through a channel the interviewer can't see: a hidden overlay, a second device, or an audio feed. The differences are mostly about where the answer appears, and that single variable turns out to determine whether any given detection approach can see it at all.

The scale is no longer hypothetical

The hard data has caught up with the anecdotes. In its 2025 AI in Hiring Report, Greenhouse found that 65% of hiring managers have caught applicants using AI deceptively. Resume Genius reported in 2026 that 22% of job seekers already use AI during real-time interviews. One detection vendor, analyzing nearly 20,000 interviews, reported that the share of flagged interviews more than doubled in the back half of 2025.

These numbers measure different things — admitted use, detected deception, algorithmic flagging — and the true underlying rate is genuinely unknown. But the direction isn't in dispute. Every credible source points the same way: up, fast.

It's also a security problem now, not just a hiring-quality problem

The most serious development isn't about candidates inflating their skills. It's about who's on the other end of the call. Through 2024 and 2025, a wave of reporting established that fake-candidate schemes had reached even security-sophisticated organizations. The security-awareness firm KnowBe4 disclosed it had unknowingly hired a North Korean operative who passed four video interviews using a stolen identity. Amazon's security chief reported blocking more than 1,800 suspected North Korean operatives from job processes since early 2024. The U.S. Department of Justice, in a late-2025 action, tied these schemes to more than 136 victim companies. Gartner projects that by 2028, one in four candidate profiles worldwide will be fake.

The implication for hiring teams is sharp: a fraudulent hire is no longer just a wasted salary. It can mean malware on a corporate laptop on day one, or a sanctioned foreign operative inside your systems. Interview integrity has quietly become a board-level control.

Why screen sharing and proctoring don't close the gap

Here's the part that catches most teams off guard. The obvious defenses — ask the candidate to share their screen, or use a proctoring tool — were designed for a different threat.

Screen sharing shows you what the candidate chooses to share. The current generation of AI assistants renders answers on a layer engineered to be absent from exactly that share. The interviewer sees a clean screen; the candidate sees the answers. Screen sharing is structurally unable to show what it was built not to capture.

Proctoring tools, meanwhile, were built largely for exams. They watch the webcam, scan the room, lock down the browser, and flag suspicious behavior. That's useful against a person looking away or a note on the wall. It's far less useful against a tool that runs quietly on the candidate's own machine, shows nothing on the webcam, and triggers no obvious behavior. And the heavier these tools get, the more they raise a second problem — fairness — that the proctoring industry learned about the hard way through years of bias and accessibility complaints.

The gap is architectural, not a tuning problem. A candidate can keep the interview tab focused, avoid copy-paste, hold steady eye contact, and still be reading AI-generated answers the whole time.

What actually closes it

Closing this gap takes two things working together, because the threat now lives in two different places.

For tools running on the candidate's machine, you need detection that identifies the tool itself and proves it was there — not behavioral guesswork, and not a list of known program names that a renamed or disguised tool walks right past. The detection has to target what the whole category depends on, so it still works against a tool that didn't exist yesterday.

For tools that deliberately run off the machine — a second phone, a tablet, an earpiece — no on-machine software can see them, and any honest vendor will tell you so. Those are closed not by detection but by how the interview itself is run: a visual-first questioning approach that denies an off-device assistant the context it needs to keep up.

This is the model Capifiq is built on: deterministic detection of what's on the machine, paired with an interviewer methodology for what isn't. On the machine, prove it; off the machine, defeat it.

The takeaway for hiring teams

The question is no longer whether candidates can use AI during interviews — they can, the tools are abundant and cheap, and adoption climbs every quarter. The question is whether your process can tell an authentic performance from an assisted one, and a real person from a fabricated one. Most remote hiring processes today can't, and the tools designed to exploit that are getting better every month.

You don't have to take anyone's word for how big the gap is. The most direct way to see it is to run real interviews through a system built for this threat and look at what it catches that your current process misses.

Capifiq runs alongside your existing Zoom, Teams, and Meet interviews and flags hidden AI tools in real time, with timestamped evidence. The first five interviews are free.