How common is candidate fraud in 2026?

Widespread. Checkr's survey of 3,000 US hiring managers found 86% of organizations experienced at least one type of hiring fraud in the past year. 40% had interviewed a fake-identity candidate, and 67% encountered an interview proxy. The problem is not limited to tech. It spans industries, roles and seniority levels.

Can AI detect deepfake candidates?

Dedicated deepfake-detection tools exist, but no single tool is definitive. Human accuracy at spotting deepfakes averages about 55%, barely above chance. The more practical approach is layered verification: structured screening that tests real knowledge, live identity checks, proctored video, and post-hire performance monitoring. Each layer catches what the others miss.

Should we stop using video interviews because of deepfake risk?

No. Video interviews are valuable precisely because they capture signal that a resume and phone screen cannot: communication ability, spontaneity, domain knowledge in real time. The answer is to add integrity controls to video (proctoring, live ID verification, structured questioning) rather than to abandon the format. Companies that have dropped video in response to deepfake fears have shifted the advantage to candidates who are willing to cheat in other ways.

Is it legal to use AI screening to detect fraud?

AI screening evaluates candidates against job-related criteria. That is legal in every US jurisdiction, subject to the same equal-opportunity and adverse-impact requirements as any other screening method. Proctoring during assessments is also standard practice. What varies by jurisdiction is the disclosure requirement: several US states and the EU AI Act require employers to tell candidates when AI is used in hiring decisions. Check your state's specific AI hiring disclosure law and make sure candidates are informed before the screening begins.

What is an interview proxy and how do you catch one?

An interview proxy is someone who takes the interview on behalf of the actual candidate, either by appearing on camera in their place or by feeding answers off-screen while the candidate lip-syncs or relays them. The most reliable detection methods are: requiring camera-on with a live ID hold at the start of the interview, using proctored video that monitors face consistency and gaze patterns throughout, and adding a live technical walkthrough where the candidate explains work they supposedly completed. Proxies can rehearse answers. They cannot explain reasoning they were not part of.

How does structured screening reduce resume fraud?

An AI-written resume passes an ATS keyword filter because it was designed to. Structured screening shifts the evaluation from what the resume says to what the candidate can demonstrate. When the screening asks a candidate to describe how they handled a specific situation, walk through a technical decision, or explain a methodology they claim to know, the gap between a fabricated resume and the person behind it becomes visible. The screening report documents the candidate's actual responses, giving the hiring team evidence to compare against resume claims.

86% of Companies Hit by Hiring Fraud — Detect Fake Candidates Before They Start

A recruiter at a mid-sized staffing firm screens a software engineer's resume. Solid GitHub, clean LinkedIn, three years at a recognizable company. The video interview goes well. Composed, articulate, technically fluent. The candidate starts on Monday. By Wednesday, the hiring manager notices the person on Zoom doesn't match the person on the badge photo, and the code submissions look like they were written by someone else. The firm has onboarded a proxy, someone who interviewed on behalf of the person who showed up to work. This is not a hypothetical. It happened to 29% of US business managers surveyed in 2026.

Candidate fraud is no longer an edge case. AI-written resumes, deepfake video interviews and professional proxy networks have turned it into an operational problem that affects hiring at every level. This article covers what the data says, which types of fraud you are most likely to encounter, what actually works for detection, and where structured screening fits into a layered defense. Eximius is an AI screening platform we build, so read those references with that in mind.

The scale of the problem

The numbers have moved fast. Checkr's 2026 Hiring Hoax survey of 3,000 US hiring managers found that 86% of organizations experienced at least one type of hiring fraud in the past year. 40% of managers personally interviewed a candidate later revealed to be using a fake identity, up from 31% in 2025. And 67% encountered an interview proxy, someone answering on behalf of the actual candidate, up from 35% the year before.

The pipeline is getting noisier, too. Pin's 2026 recruiter survey found that 61% of recruiters now see fake or AI-generated applications regularly, with nearly a quarter saying it happens constantly. An analysis of 19,368 live interviews by Fabric flagged 38.5% of candidates for AI-cheating behavior between July 2025 and January 2026, up from 9% at the start of that window. Gartner projects that one in four candidate profiles worldwide will be fake by 2028.

The cost is real but uneven. Checkr found that 15% of managers estimated hiring-fraud losses above $50,000 in the past year, and 7% above $100,000. At the same time, 51% reported minimal losses, under $10,000 or none at all. The damage depends on how quickly you catch it. Organizations that detect fraud within the first week contain the cost. Those that don't can face months of salary, access exposure and the disruption of re-hiring the role.

Three types of fraud that matter most right now

1. AI-written resumes and applications

This is the highest-volume problem. A candidate uses ChatGPT or a similar tool to generate a resume tailored to your job description, inflating skills, fabricating project experience, or rewording the JD itself into a personal summary. A Software Finder survey of 874 HR professionals found that 72% had encountered AI-fabricated resumes, portfolios or credentials. The resumes parse perfectly through ATS keyword filters because they were written to do exactly that.

The risk is not that a candidate used AI to polish their language. That is now normal and arguably reasonable. The risk is that the resume describes capabilities the person does not have, and your screening process has no way to verify them before the interview stage.

2. Interview proxies and real-time AI assistance

Interview proxies are not new, but the tools have gotten better. Hidden AI overlays like Cluely and Interview Coder feed answers to candidates during live video calls, invisible to standard screen-sharing detection. Karat estimates 80% of candidates use large language models during code tests, even when explicitly banned. And Checkr's survey found 67% of hiring managers have encountered a proxy situation, where someone other than the candidate answers the interview questions.

The structural problem is that remote video interviews removed the one verification step most hiring processes relied on: being in the same room as the candidate. Without a replacement, the process trusts that the person on camera is the person who applied.

3. Deepfake video and audio

Deepfake interviews are the most visible fraud type in the press and the hardest to detect. Pin's survey found that 22.4% of recruiters confirmed a deepfake or identity swap in a video interview, and another 16.3% suspected one but couldn't prove it. Pindrop's Voice Intelligence Report showed deepfake fraud attempts in hiring rose 1,300% between 2023 and 2024. Human accuracy at detecting deepfakes averages 55.54% across 56 studies. That is barely better than a coin flip.

This is also where state-linked threats enter the picture. The FBI warned in October 2026 that North Korea operates thousands of state-controlled remote IT workers, increasingly using US-based facilitators. The DOJ indicted 14 North Korean nationals in December 2024 over a scheme that allegedly moved $88 million across more than 100 US companies over six years.

What actually works for detection

No single tool catches every type of fraud. The organizations that contain the problem treat it as a layered defense, not a checkbox.

Structured screening that tests the person, not the resume

The single most effective counter to AI-written resumes is to stop relying on the resume as your primary evaluation signal. A structured screening conversation, where the candidate answers role-specific questions in their own words in real time, tests whether they can demonstrate the capabilities the resume claims. If the resume says "led migration from monolith to microservices," the screening asks them to describe the decision points, trade-offs and outcomes. AI-polished resumes pass keyword filters. They do not pass structured follow-up questions about work the candidate did not actually do.

This applies across modalities. Chat-based screening catches candidates who struggle to articulate what their resume describes. Voice screening adds a layer of spontaneity that prepared scripts and AI overlays find harder to handle. Video screening with proctoring adds identity-level verification to the evaluation by monitoring for tab switching, secondary devices, gaze patterns, or face mismatches.

Identity verification at multiple stages

Checkr's data shows 29% of organizations verify identity at application, 49% after at least one live interview, and 18% not until the final round or later. The FBI recommends what they call "identity in depth," which means verification at several points across the employment lifecycle, not just once. At minimum: government ID check at application, a live identity confirmation at interview with camera on and ID held to camera, and a document re-check at onboarding. The cost of three checks is trivial compared to onboarding a proxy.

In-person rounds for final evaluation

In-person interview requests rose from 5% of roles in 2024 to 30% in 2025. Companies including Google, McKinsey, Deloitte and Cisco have reportedly reinstated in-person rounds specifically to counter AI fraud. This is not a return to pre-remote hiring. It is a single verification gate, usually the final one, that confirms the person who interviewed is the person who will work. For remote roles, a single in-person onboarding day serves the same purpose.

Technical assessments with integrity controls

CodeSignal reported that technical assessment cheating rose from 16% to 35% of attempts. Proctored assessments with browser lockdown, camera monitoring and behavioral analysis reduce this, but they do not eliminate it. The better approach is to combine an assessment with a live walkthrough: the candidate completes a take-home or timed challenge, then explains their approach on a call. Someone who used AI to generate the code cannot explain the reasoning behind it.

Post-hire monitoring

Checkr found that 44% of organizations lack a formal, written process for post-hire identity verification. HYPR's research found that 42% of fraudulent candidates passed pre-hire screening and started the job, with an average of 5.73 days of unmonitored network access before detection. A 30, 60 and 90-day performance check against the screening data catches fraud that slips through the front door. Does this person's work match what they demonstrated in the evaluation? If not, you want to know at day 30, not day 180.

Where Eximius fits in a fraud defense

Eximius is not a fraud detection platform. It is an AI screening tool, and structured screening is one layer of the defense described above.

Sia conducts structured screening conversations over chat, voice or video against criteria the recruiter sets. The output is a screening report with a per-dimension evaluation of what the candidate said, including the questions asked, the responses recorded and a written rationale for each score. Not a keyword match. This makes it harder for a candidate to coast on an AI-written resume alone. The resume gets them into the screening, but the screening tests whether they can back it up.

For video screenings, Eximius offers proctoring that generates a separate proctoring report, adding an integrity layer to the conversation. The screening report and proctoring report together give the hiring team both evaluation data and a documented audit trail of how the candidate engaged.

What Eximius does not do: it does not perform government ID verification, deepfake detection, or background checks. Those steps sit upstream and downstream of the screening phase and require dedicated identity-verification and background-check providers. A layered approach means Eximius handles the structured evaluation, an identity provider handles the ID check, and a background-check vendor handles the credential and history verification.

Compliance: SOC 2, ISO 27001, SAML/SSO. Pricing is credit-based and public: individual plans from $100/mo, business plans from $300/mo to $2,000/mo, with chat screening at 2 credits, voice at 5 and video at 10.

A fraud-aware hiring checklist

Five steps to reduce fraud exposure without slowing your hiring process to a crawl:

Screen the person, not the document. Use structured screening, whether human or AI-conducted, that asks candidates to demonstrate capabilities, not just list them. This is the single highest-leverage change for teams currently screening on resumes alone.
Verify identity before you invest interviewer time. A government ID check at application plus a live camera-and-ID confirmation at the first video interview catches most identity mismatches before the panel round.
Add proctoring to video evaluations. If your screening or assessment includes a video component, enable proctoring. The cost is marginal and the signal from gaze patterns, tab activity and face consistency catches the most common proxy and AI-assistance methods.
Separate the assessment from the explanation. For technical roles, have the candidate complete an exercise and then walk through their approach live. This is the most reliable way to distinguish someone who did the work from someone whose AI did.
Build a post-hire verification window. Set a 30-day review point where the hiring manager compares on-the-job performance against screening data. Document the comparison. If something doesn't match, you want to know at day 30, not day 180.

For more on how AI screening works in practice, see our guide to what AI candidate screening actually does and doesn't do. For the regulatory landscape around AI in hiring, read our overview of 2027 AI hiring law obligations for screening teams.

Want to see how structured screening and proctored video work on your open roles? Book a demo and we'll run it through the Eximius workflow.