Candidate Fraud in Hiring: Catching Proxy Interviews and Faked Resumes
Remote interviews and AI tools have made it easier for someone to fake their way into a role. Here's where the fraud actually gets through, and how to build checks that don't scare off real candidates.
Candidate fraud now includes proxy interviewees, AI-assisted live answers, and fabricated work histories that slide past standard checks. Catch it by verifying identity on camera before the final round, keeping one unscripted question in every loop, calling references by phone instead of accepting written letters, and confirming at least one employer directly rather than relying on a database match.
Why this looks different than it did five years ago
A decade ago, candidate fraud mostly meant an inflated title or a stretched date range on a resume, the kind of thing a diligent background check would catch. The interview loop itself was considered safe, because the person on the call was, by definition, the person who'd show up for work.
That assumption doesn't hold anymore. Interview loops are conducted entirely over video by default. Technical assessments and writing samples are done unsupervised, on the candidate's own time, on the candidate's own machine. A convincing professional profile can be built in an afternoon. And generative AI tools can feed a candidate real-time, well-phrased answers through a second screen or an earpiece without the interviewer ever knowing. The fraud has moved from 'lying about the past' to 'substituting identity in the present' — a materially harder thing to catch after the fact.
The patterns hiring teams are actually running into
Proxy interviewing is the most serious version: a more skilled person interviews under the candidate's name and credentials, then the less-qualified person who actually got hired shows up on day one. This shows up most in roles where competency is judged through a single technical conversation rather than observed work — data, analyst, and engineering-adjacent professional roles are frequent targets.
AI-fed answers are more common and subtler. The tell isn't that the answer is good, it's the rhythm: a slight lag before each response, delivery that sounds read rather than reasoned, and polish that's identical in a scenario question as it is in a scripted 'tell me about yourself' — real candidates are rarely equally smooth at both.
Fabricated continuity is the old fraud wearing new clothes: a resume with no visible gaps, but dates that don't line up with what the employer's own records would show, or a title that got promoted a level during the application process. Reference fraud rounds it out — a friend or relative presented as a 'former manager,' or a real former colleague who will say anything to help the candidate land the job, regardless of what the work was actually like.
Where your current process already leaks
Most of the exposure isn't about having bad instincts, it's about structural gaps that fraud has learned to exploit. If identity is never verified against a photo ID until a background check runs post-offer, there's an open seam between 'who interviewed' and 'who was hired' that nothing in the loop actually closes.
If reference checks happen by email or a web form instead of a live phone call, you lose tone, hesitation, and the ability to ask a question the reference didn't prepare for — which is most of what makes a reference check useful in the first place. And if a background-check vendor confirms employment dates and titles purely through a database match rather than a direct call to the former employer, an inflated title or a padded date range often sails through untouched.
A detection framework that doesn't slow the search down
You don't need a forensic process for every hire, just a few fixed checkpoints applied consistently. Verify identity once, on camera, against a photo ID before the final round — not after the offer. Keep one unscripted, scenario-specific question in every loop that can't be prepped from the job description, and pay attention to how the answer is constructed, not just whether it's correct.
Compare the candidate's live spoken reasoning against their written work sample. A mismatch in sophistication between the two — a polished take-home assignment paired with vague, hesitant live explanation of that same work — is the single most reliable tell available to a hiring manager without specialized tools. Call references by phone and ask something they couldn't have anticipated, like what they'd have changed about how the candidate worked with their team. For roles above a meaningful compensation threshold, confirm at least one prior employer directly rather than trusting a database match alone.
What a recruiting partner should already be doing for you
If you're paying a direct-hire fee, identity and reference rigor shouldn't be something you're building from scratch on every search — it should already be baked into how candidates are sourced and vetted before a resume ever reaches you. Ask any search partner directly: do they verify identity on camera, do they call references by phone, and do they flag when the same candidate profile resurfaces across multiple searches with inconsistent details. A firm that can't answer those questions plainly is asking you to absorb risk it should be screening out.
Where to draw the line without running an interrogation
Over-scrutinizing every candidate burns real candidates out of the process and invites the same kind of fatigue covered elsewhere on interview loop design. The fix is proportionality, not suspicion: apply the same fixed checkpoints to every candidate for a given role, rather than only to the ones who happen to seem off, which protects you legally as much as operationally. Scale the rigor to the risk — a controller or finance-systems hire with access to payment rails warrants a deeper employer verification than an individual-contributor marketing role, and that calibration should be set before the search opens, not improvised mid-loop.
Frequently asked
Good questions.
How do we know if a candidate used AI during a live interview?
Watch the rhythm more than the content. A short, consistent lag before every answer, delivery that sounds read rather than reasoned, and polish that's identical whether the question is a scripted behavioral prompt or an improvised scenario are the practical tells. Asking a follow-up that requires them to defend or adjust their own prior answer is the cleanest test, since a scripted or fed response usually can't adapt on the fly the way a real, lived answer can.
Should every role require a video ID check before the final round?
It's a reasonable default rather than an overreaction, and treating it as a standard step for every candidate avoids both the seam where proxy interviewing happens and any appearance of singling someone out. Keep it brief and routine: a quick photo ID match on camera before the last interview, framed the same way for every finalist regardless of role or background.
What do we do if we discover fraud after someone has already started?
Treat it as a distinct issue from a performance-based guarantee claim, since most direct-hire guarantees are written around fit and performance, not misrepresentation discovered after start. Document exactly what was misrepresented, loop in HR and legal before taking action, and separately flag it to whatever recruiting partner sourced the candidate so they can adjust their own screening and verification process going forward.
Is proxy interviewing actually a widespread problem or an overreaction?
It's hard to quantify precisely because by design it's meant to be invisible, which is exactly why structural checkpoints matter more than trying to gauge how common it is. Rather than deciding whether to worry based on prevalence, build the identity and reference verification steps in as a fixed cost of doing remote-first hiring, the same way you'd treat a background check as routine regardless of how often it turns something up.
Ready to make the hire?
Tell us the role, the comp band, and the timeline — we'll tell you exactly how we'd run the search.