You Can't Call It Cheating If You Never Drew the Line

When was the last time you used an AI tool to handle something tedious?
For me, it was about a minute ago.

And filling out job applications is about as tedious as it gets. Customizing a resume for every role, answering the same screening questions over and over, grinding through online questionnaires. Candidates hate it. So naturally, over 50 start-ups have popped up promising to fix that, with AI tools that customize your profile and fire off applications on your behalf.

Sounds great, right?
Maybe not if you're the one hiring.

According to a CNBC report from early 2025, nearly two-thirds of candidates now use AI somewhere in their job applications. That's not a forecast. That's the baseline your pipeline already runs on. The people applying to your roles are drafting with AI, researching with AI, and prepping for your interviews with AI, whether or not you've said a word about it.

And here's what's surprising: given how easy it is for candidates to knowingly or unknowingly reshape how their experience and skills come across, most hiring teams still aren't sounding the alarm. Most haven't given candidates a single instruction about what's fair game and what isn't.

Therefore, candidates are left to figure it out themselves, and they land everywhere from cautious to bold. In a May 2025 survey of 600 US hiring managers by TopResume, 52% said it's fine to use AI to proofread or draft an application. But 40.8% said using AI in a skills assessment is unacceptable, and 57% said AI should never touch an interview. No two employers draw that line in the same place, which leaves plenty of room for candidates to get resourceful.

Everyone agrees on the obvious stuff: willful misrepresentation, lying, stealing answers from another source, that's not okay. But most candidates aren't doing that. They're enhancing. Embellishing. Turning a "B" profile into an "A" with nothing but good intentions, while employers stay busy chasing the rare, obvious offenders.

There's more than one kind of cheating

In ambiguous situations, good candidates tend to hold back. Someone who uses AI fluently every single day, exactly the skill a lot of these roles now require, switches it off during your assessment because they're worried it'll count against them. So you end up measuring who was most willing to follow a rule nobody wrote down, not who's actually best at the work. And trust is already low to begin with: Gartner found only 26% of applicants trust AI to evaluate them fairly, and half believe AI is already screening them out. Piling ambiguity on top of that doesn't make candidates more honest. It just makes them more anxious, and less like themselves.

Picture a software engineering candidate who's never actually worked with a particular programming language or cloud platform. An AI tool scans the job description, spots the exact keywords the employer wants, and rewrites the resume to sound like the candidate has that experience. The candidate probably isn't lying on purpose. The resume is just a lot more polished and targeted than they could have written on their own. But now the employer is evaluating someone who looks like they have technical skills they might not actually be able to demonstrate on day one of the job.

Same story in sales. Take a candidate applying for an enterprise account executive role with limited experience selling to large organizations. AI rewrites the resume to spotlight enterprise sales, strategic accounts, complex deal cycles, executive relationships, and helps craft sharp answers to the screening questions. If that candidate then leans on AI during the actual assessment or interview, the employer isn't evaluating their judgment or communication skills anymore. They're evaluating the output of a tool. Some candidates get the questions ahead of time and prep specifically for them. Others just copy answers from somewhere else. Different tactics, same problem: the process is supposed to measure what the candidate can actually do, and something else has slipped in and changed the signal.

While many cases involve augmenting, usually driven by a genuine desire to do well, stand out in a crowded market, and get a real chance at being seen, some of what we see is flat-out deception. Such behaviors go further than misrepresentation, into outright fraud. Real-time interview copilots feed candidates answers during live calls. Deepfake video and voice tools change who's actually on the other end of the screen. Gartner projects that by 2028, one in four candidate profiles worldwide could be fake, and in its own survey of 3,000 candidates, 6% already admitted to some form of interview fraud, like having someone else pose as them. One analysis of more than 19,000 technical interviews by the integrity firm Fabric HQ flagged 38.5% for signs of live AI assistance in the second half of 2025. Take that number with a grain of salt, since it's one vendor's flags rather than proven cheating, but the direction is hard to wave away. Guardrails matter here: verifying identity, protecting assessment content, building processes that catch deception from the start.

Yet, you can't fairly call something cheating if you never defined the rules. And you can't ask candidates to be honest about their AI use while staying silent about your own.

Most of the industry is stuck in the wrong debate

Faced with all this, hiring tech has mostly retreated into one of two corners.

Corner one bans AI outright. Candidates sign honesty pledges promising not to touch the tools, and the assessment leans on detection software to enforce it. Understandable logic, but it sets up an arms race the assessment can't win, and it quietly measures the wrong thing. Telling a candidate not to use AI on a task they'd absolutely use AI for on the job tests their compliance, not their capability.

Corner two just ignores the question and hopes the scores still mean something. Increasingly, they don't. Fabric HQ's analysis found that 61% of flagged candidates scored above the pass threshold, meaning an untouched scoreboard can quietly reward exactly the behavior you were worried about.

There's a third path, and the smartest employers are already on it. Greenhouse, one of the most widely used hiring platforms out there, now publishes plain, candidate-facing guidelines that welcome AI for prep and for work assessments, as long as the final submission reflects the candidate's own synthesis and judgment. Then it draws a firm line at reading AI answers live, in an interview. That's not permissiveness. That's precision. It tells candidates exactly what "good" looks like, which is the only thing that makes a line actually enforceable.

Decide what you're measuring, then say it out loud

This is where our advice comes in. We work with clients to communicate openly with applicants about the guardrails for an assessment: what they can use, what they can't, what preparation is fair, and what the assessment is actually designed to measure. Just as important, we want candidates to understand why those guardrails exist. Not because "the employer said so," but because the goal is a fair shot to show the capabilities they'll actually bring to the job.

When a hiring process gives candidates real guidance on AI, it shows. The strongest policies we've seen all do four things.

They welcome AI where it mirrors the job. On a skills assessment or a practical task, candidates use the same tools they'd use in the role, unless a specific task says otherwise. That keeps the assessment honest to the actual work, and rewards people who use the tools well. Which, increasingly, is the whole point.

They draw a firm, specific line at live assistance and misrepresentation. No real-time copilots in interviews. No deepfakes. No one else standing in for you. Easy to understand, easy to defend, because it protects a genuine conversation instead of policing preparation.

They're transparent in both directions. If you expect candidates to be open about how they use AI, tell them how you use it too: where it screens, where it grades, where a human reviews, and what it doesn't do. Openness earns openness.

They keep a human accountable for the decision. AI can help a team review work consistently, at scale. It shouldn't be the thing that decides who moves forward. Candidates deserve to know a person is making the call, and hiring teams need that line for their own governance, too.

Vervoe’s Guidance

We built Vervoe on a simple idea: the best evidence of whether someone can do a job is watching them actually do the work, not reading about where they've been. AI-powered screening, autograded skills assessments, and job simulations show how candidates perform on realistic, job-related tasks, graded against the same criteria for every applicant. The score is decision-ready evidence for a person to weigh. The hiring decision stays with a human, and the reasoning behind it stays explainable and auditable.

That distinction points to something worth naming outright: not every skill in a hiring process should be treated the same way. Some have to be "owned" by the candidate. Others can be "artificially acquired," and that's fine.

Say you're hiring a salesperson. Building rapport, running a product walkthrough, negotiating a deal: those aren't skills you can outsource to a tool and still call the person a salesperson. They're non-negotiable. Owned. Which means they need to be tested in a way that can't be gamed: candidates have to actually perform the task, live, and they should know upfront exactly what resources they're allowed to lean on, if any.

That's why the AI question isn't some awkward exception for the teams that hire with us. It's the natural extension of how they already assess. If your process is built to see real skills through real work, then telling candidates to use the tools of that work, within clear limits, is just consistent. The alternative, an assessment that pretends AI doesn't exist, means measuring a world that no longer does.

AI proficiency doesn't stop mattering once someone is hired. The same evidence-first approach applies to the people already on your team, which we cover in how to measure AI readiness across your workforce.

Candidates have already decided. The only open question left is whether hiring will define the line before it starts calling people cheaters for crossing one that was never there.

Sources

Gartner, candidate fraud and fake profiles research, released July 31, 2025 (survey of 3,000 candidates), as reported by HR Dive.

TopResume survey of 600 US hiring managers, May 2025, via Pollfish: Where employers draw the line on AI in hiring.

CNBC, February 2025: Nearly two-thirds of job candidates are using AI in their applications.

Fabric HQ, analysis of 19,368 interviews (July 2025 to January 2026), published January 14, 2026: State of AI interview cheating. Vendor dataset; figures reflect flagged behavior.

Greenhouse, published candidate policy: Guidelines for using AI in our interviewing process.