Your Next AE Interviewed With a Copilot: How AI Broke Sales Hiring and What to Test Instead

The candidate was great. Crisp on discovery, calm on objection handling, a clean answer to "walk me through a deal you lost." Your panel scored him a strong yes. Four months later he can't run a first call without a script, and the manager who hired him is quietly asking whether he was the same person who interviewed.

That story used to be rare enough to tell as a war story. It isn't rare anymore.

Hiring has turned into an arms race between two sets of AI tools. Candidates use them to write resumes, beat screening filters, and in a growing number of cases feed themselves answers during live interviews. Employers use them to sort the flood that results. Greenhouse's CEO, Daniel Chait, called it an "AI doom loop" when his company published its 2025 AI in Hiring Report, and the phrase fits.

For CROs, VPs of Sales, frontline sales managers, and RevOps leaders who own headcount plans: the interview loop you built in 2021 was designed to spot people who can sell. It was not designed to tell a seller apart from a seller with an invisible assistant, or from someone who isn't who they claim to be. With ramp times at a record high, every hire that slips through costs more than it used to. This article looks at what the research says about AI in the hiring funnel, why sales roles carry a specific version of the risk, and how to redesign the loop so you test the things AI can't fake for the candidate.

The numbers behind the doom loop

Start with volume and trust, because they explain everything else.

Greenhouse surveyed more than 4,100 job seekers, recruiters, and hiring managers across the U.S., U.K., Ireland, and Germany for its 2025 report. In the U.S., 91% of recruiters said they had spotted candidate deception. A third (34%) said they spend up to half their week filtering spam and junk applications. Two in three hiring managers (65%) had caught applicants using AI deceptively: 32% had seen candidates reading from AI-generated scripts, 22% had found prompt injections hidden in resumes, and 18% had encountered deepfakes. And 74% of hiring managers said they were more worried about fake credentials, deepfakes, or misrepresented experience than they had been a year earlier.

The candidate side of that same survey is just as striking. Among 1,200 U.S. job seekers, 41% admitted to using prompt injections (hidden text meant to steer an AI screener), and more than a third (36%) said they had used AI to alter their appearance, voice, or background during a video interview. Most of those people are not fraudsters. They're anxious applicants who believe an opaque algorithm is filtering them out, so they fight algorithm with algorithm. Only 8% of job seekers told Greenhouse they think AI makes hiring fairer.

Gartner's research points the same direction. In a survey of 3,000 job candidates, 6% admitted to participating in interview fraud, either posing as someone else or having someone else pose as them. Four in ten said they use AI during the application process, mostly to write resume text, cover letters, writing samples, or answers to assessment questions. Gartner's forecast from that work is the one that gets quoted in board meetings: by 2028, one in four candidate profiles worldwide will be fake.

If that sounds like a prediction for later, GetReal Security's Deepfake Readiness Benchmark Report suggests later is already here. Among 668 IT, security, risk, and fraud leaders at companies with at least 1,000 employees, 41% said their company had hired and onboarded a fraudulent candidate.

What cheating in a live interview looks like now

The interview stage is where this gets uncomfortable for sales leaders, because it's the stage we trust most.

Fabric, an AI interviewing platform, analyzed 19,368 interviews run on its system between July 2025 and January 2026, each with automated cheating detection. It flagged 38.5% of candidates for cheating behavior. The rate roughly tripled between July and September 2025, which lines up with a new generation of tools that put invisible overlays on the candidate's screen. The interviewer sees a normal screen share. The candidate sees a live teleprompter.

Fabric broke down the methods it caught. Dedicated cheating assistants accounted for 45% of cases, voice mode on general chatbots like ChatGPT for 34%, old-fashioned tab switching or a second screen for 18%, and live help from another human for just 3%. The two leading methods, 79% of the total, are built to be invisible to screen sharing.

Then comes the finding that should bother anyone who runs a scorecard-based panel: 61% of flagged cheaters scored above the pass threshold. The assisted answers weren't just passable. They were the textbook answers interviewers are trained to reward, delivered without the hesitation that normally costs a candidate points.

For sales, the picture is different. Fabric found technical roles cheating at 48% and sales roles at 12%, roughly a quarter of the rate. Fabric's explanation is that sales interviews lean on open-ended, soft-skill questions that are harder to game than coding problems with one right answer.

I'd read that as good news with an asterisk. Twelve percent is still more than one candidate in ten. And the reason sales scores lower is the same reason a lot of sales loops feel safe: they rely on conversation. The moment your loop leans on predictable questions ("tell me about your sales process," "how do you handle a price objection"), it starts to look like the kind of interview AI handles well. A polished, well-structured answer to a common question is now the cheapest thing in the world to produce.

Why a bad sales hire costs more in 2026

Every hiring mistake is expensive. The Bridge Group's 2026 AE research makes the sales version unusually easy to quantify.

The study, the tenth edition, covering 158 B2B companies, puts average AE ramp time at 6.2 months, the highest in the study's history. Median AE on-target earnings rose to $200,000. Median quota is $960,000. Only 48% of reps are hitting quota, down from 51% in 2024.

Do the rough math on a mis-hire. Six months of ramp on a $200,000 OTE is on the order of $100,000 in compensation before a rep is expected to be fully productive, and that's before recruiting fees, manager time, and the pipeline the seat was supposed to produce. If the hire washes out, you pay for the backfill and start the ramp clock again. With fewer than half of reps reaching quota even in a normal year, you have very little slack for seats filled by people who interviewed better than they sell.

The Bridge Group data holds a second clue. Companies now require an average of 3.7 years of prior experience for AE hires, up from 2.7 years in 2022, and the report says the era of the junior AE "appears largely over" at higher deal sizes. Fabric found junior candidates (zero to five years of experience) cheating at nearly double the rate of senior ones. When the bar rises and the entry-level market gets tighter, the incentive to inflate experience grows with it. Your job description asks for four years of enterprise selling, and some share of applicants will make sure their resume says exactly that.

The paradox: you want reps who are good with AI

This is where the conversation usually goes wrong. The easy reaction is to treat any AI use by candidates as cheating and try to ban it. That misses what you're hiring for.

The same Bridge Group study found that organizations in the top third for AI engagement had 57% of reps at quota, compared with 39% in the bottom third. The researchers are careful to call this observational (companies that invest in AI may differ in other ways), but the direction is hard to ignore. Post-call summaries, account research, and email personalization were the most common AI-assisted AE workflows. You want the person you hire to be fluent in exactly that kind of work.

So the problem isn't that candidates use AI. The problem is misrepresentation: an interview that measures the tool while you think you're measuring the person. Fabric put it bluntly in its report, describing the risk as hiring "a subscription that mimics a human."

Once you frame it that way, the design goal gets clearer. A modern sales interview loop has to measure two separate things: how well the candidate works with AI when AI is allowed, and how well they think and sell when it isn't there. Most loops measure neither on purpose. They measure a blend and hope.

Redesigning the sales interview loop

None of what follows requires new software, though detection tools exist and some teams will want them. It mostly requires deciding in advance what each stage is for.

Publish your AI policy before the first screen

Greenhouse's 2026 Candidate AI Interview Report found that 63% of active job seekers have now been interviewed by an AI, yet 70% were never clearly told upfront that AI would be evaluating them, and 21% only found out once the interview started. Candidates are guessing about your rules and assuming everyone else is breaking them. Fabric describes this as a prisoner's dilemma, where honest candidates cheat because they assume the competition already is.

Write down what's allowed at each stage and send it with the interview invite. For example: AI is fine for preparing your research on our company, fine and encouraged in the take-home, and not allowed during live conversations. Gartner recommends the same thing as part of a layered fraud strategy: set clear expectations for acceptable AI use and tell candidates about your fraud detection efforts. Clear rules cut the honest candidates' reasons to cheat and make violations easier to act on.

Make one stage AI-open and grade the thinking, not the output

Take the classic take-home (a territory plan, an account plan, a prospecting sequence for three named accounts) and explicitly allow AI. Then change what you score. Everyone's document will look good now, so the document is not the signal.

The signal comes in the debrief. Have the candidate walk you through the plan live and ask why. Why these three accounts and not the next three? Which of the AI's suggestions did you throw out, and what was wrong with them? What would you check before sending this email to a real CFO? A rep who used AI well can tell you where it was wrong. A rep who pasted the output can't, and it shows within about five minutes.

This stage also tells you something your old loop never did: whether the candidate already works the way your top-quartile reps do.

Make one stage live, unassisted, and unpredictable

Role-plays are the backbone of most sales loops, and they're more resistant to cheating than Q&A. They're still gameable if they're predictable. A standard discovery role-play with a cooperative buyer and a stock objection is exactly what an assistant can script.

Break the pattern. Have your "buyer" change the scenario halfway through: the champion just left, procurement wants a 30% discount by Friday, the prospect reveals a competitor is already in legal. Ask follow-ups that require personal memory rather than general knowledge, like the name of the deal that taught them this lesson, what the buyer actually said, and what they'd do differently. Fabric's detection advice points the same way. When an answer sounds polished, drill into a specific failure. Genuine experience holds up under follow-ups. Generated experience goes generic.

Watch for the tells Fabric describes as well. Assisted candidates tend to show a consistent delay of a few seconds after every question regardless of difficulty, and their eyes move in steady, left-to-right reading lines while they're supposedly thinking. Neither proves anything on its own. Both are reasons to ask a harder follow-up.

Bring back one in-person moment

Some of the biggest employers have already gone this way. Google CEO Sundar Pichai said the company is reintroducing at least one round of in-person interviews, and McKinsey asks hiring managers to schedule at least one in-person meeting before extending an offer. Cisco has also spoken publicly about fake candidates getting into the hiring process. In Greenhouse's U.S. data, 39% of hiring managers said they're running more in-person interviews to verify authenticity.

Candidates mostly don't mind. In Gartner's research, 62% of candidates said they were more likely to apply to a role that required in-person interviews. For distributed sales teams, this doesn't have to mean flying every finalist to headquarters. A final-round coffee with a regional manager or a ride-along with a current rep does the job, and it doubles as a better culture signal than any video call.

Verify identity and history like you'd verify a deal

Sales leaders are trained to validate what buyers tell them. Apply the same discipline to candidates. Call references you found yourself, not only the ones provided. Ask for W-2s or commission statements to confirm attainment claims when your legal team allows it. Check that the person on the final call is the person from the first call. Gartner recommends tighter background checks, identity verification, and anomaly alerts in recruiting systems, and GetReal's 41% figure is a reminder that this is a security problem as much as a talent one. A fraudulent hire gets access to your CRM, your customer list, and your pricing.

There's a sales-specific version of backchanneling worth adding. Ask a customer the candidate claims to have closed. If the rep really ran a six-figure deal, someone on the buying side remembers them.

Reduce the flood at the top of the funnel

Much of the cheating pressure starts with volume. When recruiters spend a third of their week filtering junk, they lean harder on automated screening, candidates respond with prompt injections, and the loop tightens. You can't fix the whole market, but you can make your own funnel less attractive to mass applicants. Ask one specific, short question in the application that requires knowledge of your market (for example, "which of our competitors would you expect to face most often in the mid-market, and why?"). Referrals from your own reps, who stake their reputation on the recommendation, are also worth more than they were three years ago.

What this means for the people doing the hiring

The practical cost of all this falls on frontline managers. Greenhouse found that 68% of U.S. hiring managers are more involved in hiring than they were the year before, which runs against the promise that AI would save them time. The loop above asks managers to build better role-plays and show up in person more often.

I don't think there's a way around that, and I'm not sure there should be. Hiring a seller is a bet of well over $100,000 and six months of runway on one person's judgment. It deserves more of a manager's attention than a panel scorecard and a gut feeling. The encouraging part is that the same skills that catch a coached candidate (asking the second and third question, noticing when an answer is too smooth, verifying claims instead of accepting them) are the skills good managers already use on deal reviews.

A few measures will tell you whether the new loop is working. Track the share of new hires who reach ramp milestones on schedule, split by which interview stages they performed best in. If candidates who shone in the AI-open debrief ramp faster than those who shone in the standard Q&A, you've learned something about what predicts performance on your team. Track early attrition within the first nine months, since that's where mis-hires show up. And ask your recent hires, anonymously, what they thought of the process. Candidates who felt the rules were clear and fair are the ones who will refer the next strong seller.

Conclusion

The sales interview was always a performance. Candidates rehearsed, managers knew it, and the best interviewers learned to push past the rehearsed answer to find the real one. What changed is the quality of the rehearsal. An invisible assistant can now deliver the perfect answer to a predictable question in real time, and more than six in ten assisted candidates in Fabric's data cleared the bar.

You can't out-detect every tool, and you shouldn't try to hire people who avoid AI. The Bridge Group's data suggests the reps who use it heavily are the ones hitting quota. What you can do is stop letting one blended interview answer two different questions. Let candidates use AI where you'd expect them to use it on the job, and grade their judgment about it. Take it away where the job depends on live thinking, and make those moments unpredictable enough that a script can't keep up. Then verify what they told you with the same skepticism you'd bring to a forecast call.

Ramp is at 6.2 months, fewer than half of reps hit their number, and Gartner expects one in four candidate profiles to be fake by 2028. Loops built for 2021 will keep producing hires like the one in the opening story. The teams that rebuild theirs now will spend the next few years hiring sellers who can actually sell.