How to Spot Candidate Fraud Before It Costs You a Bad Hire

Candidate fraud used to mean a padded resume or an inflated job title. Today, the problem has evolved into something harder to detect and significantly more expensive when it slips through.

Recruiters are now dealing with AI-generated resumes that are nearly indistinguishable from real ones. They’re encountering candidates who use proxy interviewers for technical screens, swapping in a more skilled person for the video call while the actual candidate shows up on day one. And they’re seeing an uptick in what the industry calls “spray and apply” bots, automated tools that submit hundreds of applications per hour using scraped or fabricated profiles.

The cost of a bad hire has always been high. By most estimates, replacing a mis-hire costs 50% to 200% of their annual salary once you factor in onboarding, lost productivity, and the time spent re-hiring. But a fraudulent hire is worse because the failure isn’t a skill gap you can coach around. The person who shows up literally isn’t the person you evaluated.

So how are recruiting teams actually catching this?

Your ATS can do more than you think

The first line of defense is metadata, and modern applicant tracking systems are getting smarter about flagging it. Ashby, for example, now includes native fraud detection signals that look at things like the candidate’s IP location relative to their stated address, resume formatting patterns that suggest AI generation, and the age of the email address used to apply. An email created the same week as the application is a red flag worth investigating.

These signals aren’t foolproof on their own, but they create a useful triage layer. Instead of manually reviewing every application for signs of fraud, recruiters can focus their attention on the flagged subset and spend their time where it actually matters.

Greenhouse offers some of these capabilities too, but they’re typically locked behind paid add-ons that can run $10K or more per year. For smaller teams, that price tag means fraud detection often gets deprioritized until someone gets burned.

The real gap is at the live screening stage

No tool on the market has reliably solved fraud detection during a phone screen or video interview. AI can process applications and flag anomalies in structured data, but the moment a human conversation starts, the detection problem gets much harder.

Recruiters who deal with high-volume pipelines have developed manual tactics that are surprisingly effective. One approach that came up in a recent conversation among TA leaders is to intentionally misstate a candidate detail during the screen. Mention the wrong city they listed on their resume, for example, or reference a company they didn’t work at. A real candidate corrects you immediately and naturally. A fraudulent one, whether it’s a proxy or someone working from a fabricated background, tends to either agree with the incorrect detail or hesitate in a way that feels off.

Another tactic is front-loading application questions that require candidates to confirm specific details from the job description. This filters out bots and mass-appliers who never read the posting, and it creates a reference point you can return to during the screen. If someone confirmed in their application that they have experience with a specific framework and then can’t speak to it on the call, that’s a clear signal.

Recording screening calls is also becoming standard practice among teams that take fraud seriously. Tools like Evidence let recruiters document the conversation so that inconsistencies can be reviewed after the fact, shared with hiring managers, and used as evidence if a candidate is caught misrepresenting their background. One recruiter recently shared that they caught a candidate claiming to work at Meta who was actually at a much smaller company, and the call recording was the proof that made the case clear.

Customer-facing roles need a different approach

The sensitivity around AI screening is highest for roles where relationship-building is the core skill. Account executives, sales development reps, and customer success managers are all positions where the candidate’s ability to communicate and build rapport is the thing you’re evaluating. Outsourcing that evaluation to an AI screener feels contradictory, and candidates agree.

One recruiter trialed RightHire, an AI phone screening tool, for an SDR role. Out of six or seven candidates routed through it, only one completed the screen. The rest dropped off entirely. They weren’t willing to have their first interaction with a potential employer be a conversation with a bot.

For high-volume technical roles where the initial screen is mostly about confirming skills and logistics, AI screening can save real time. But for roles where interpersonal dynamics are the point, candidates are voting with their feet. They want to talk to a person, and forcing them through an AI gate first costs you the people who have the most options.

Reference checks are an underused fraud detection tool

Most companies treat reference checks as a formality, a box to tick before extending an offer. But when done well, they’re one of the best ways to verify that the person you interviewed is who they say they are.

The best reference question that surfaced in a recent recruiter roundtable was deceptively simple. Ask for “handling tips.” It sounds supportive and managers are usually happy to share. But the answers consistently reveal personality dynamics, management challenges, and working style friction that a standard “would you hire them again” question never surfaces. If a reference says something like “they need a lot of positive reinforcement” or “they work best when they feel like the smartest person in the room,” that’s real signal about what you’re getting.

For sales hires specifically, skip the soft questions and ask directly whether the candidate hit quota in year one and year two. References will dance around this if you let them, so don’t. And pay attention to tenure. Under two years at an enterprise sales role is a pattern worth investigating, because it often means the candidate ramped, had one bad year, and left before the numbers caught up.

Back-channeling is where the real intel lives

Formal references are people the candidate chose. They’re going to say nice things. Back-channeling, which means reaching out to mutual connections or former colleagues the candidate didn’t list, is where you find out what the references wouldn’t say.

This is standard practice for VP and executive hires, but it’s increasingly common for senior individual contributor roles too. The challenge is finding the right people to talk to quickly enough that it doesn’t stall the process.

Juicebox is developing a network feature that identifies second-degree connections between your team and the candidate, which could make ad hoc back-channeling much faster. And LinkedIn, for all its faults as a sourcing tool, remains the best way to identify mutual connections and request a quick off-the-record conversation.

The legal landscape is worth knowing

California and Florida both restrict the use of AI in final hiring decisions. You can use AI to flag risks, rank candidates, and generate shortlists, but a human has to make the final call. This is worth building into your process explicitly, not just as a compliance measure but as a safeguard against the kinds of errors that automated systems introduce.

The companies getting this right are the ones who treat AI as the prep layer and humans as the decision layer. AI processes the volume, surfaces the signals, and compresses the timeline. Humans evaluate judgment, verify identity, and make the call. That separation isn’t just legally defensible. It’s better hiring.

*Fonzi connects pre-vetted software engineers with startups and tech companies, so the screening starts before you ever see a candidate. If you’re hiring engineers and want to work with people who’ve already been validated, schedule a call.*