Interview with Husnain Bajwa, SVP – Risk Solutions at SEON
1. Why are employers seeing more fake candidates and resume fraud now?
Resume fraud has always existed, but the tools have changed. Generative AI lets a fraudulent applicant produce a tailored resume, a polished cover letter and convincing assessment responses in seconds. What used to take hours of manual effort now scales instantly.
The volume problem makes it worse. When recruiters are processing hundreds of applications per role during peak hiring periods, individual scrutiny drops. Patterns that might stand out in a smaller applicant pool, like a cluster of submissions arriving within minutes or multiple resumes using nearly identical language, get lost in the noise.
But the real shift is that resume fraud is increasingly tied to identity fraud. Some applicants exaggerate experience. Others use synthetic identities, proxy interviewers or stolen credentials to move through the hiring process entirely. Gartner projects that by 2028, one in four candidate profiles worldwide will be fake. According to a 2025 Deepfake Readiness Benchmark Report, 41% of IT, cybersecurity, risk and fraud leaders say their organization has already hired a fraudulent candidate.
2. What tactics are fraudulent candidates using to bypass hiring processes?
The tactics go far beyond a padded resume. At the application stage, AI generates resumes and cover letters that match job descriptions with precision. Keyword-based screening becomes less reliable because a fabricated application can look highly relevant.
At the interview stage, it gets more sophisticated. Proxy interviewers sit for candidates using real-time face-swap software. Voice cloning layers a different voice over the speaker’s actual audio. Coached responses handle standard behavioural questions. The tells are subtle with examples such as facial expressions that lag slightly behind speech, audio that falls out of sync or answers that feel rehearsed without quite matching what was on paper.
In some cases, the fraud is coordinated rather than opportunistic. Applicants build complete professional identities from scratch – fabricated names paired with AI-generated headshots, polished LinkedIn profiles, GitHub accounts populated with cloned repositories and email addresses aged for months before being attached to applications. In a 2025 survey of 3,000 hiring managers, 59% said they had suspected a candidate of using AI to misrepresent themselves, while only 19% said they were confident their current process would catch it.
3. Why are traditional hiring checks struggling to catch this kind of fraud?
Most hiring processes concentrate verification at the wrong stage. Background checks happen after interviews and sometimes even after offers. Employment verification relies on information the candidate provides. Reference calls go to contacts the candidate selects. By the time these checks run, recruiters have already invested hours in someone who may not exist.
The deeper problem is that traditional checks answer the question “is this document real?” rather than “is this person who they claim to be?” A fabricated identity built with aged email accounts, a consistent device history and a believable behavioural pattern can pass document-level checks precisely because every individual signal looks legitimate. The fraud only becomes visible when those signals are viewed together, across the full applicant pool rather than one application at a time.
4. How can organisations detect fake candidates without slowing down genuine applicants?
The answer is not adding more hoops for candidates to jump through. Legitimate applicants are already navigating hiring timelines that average 40 to 60 days. Visible verification steps risk driving away the people you actually want to hire.
What works is running detection in the background from the moment an application is submitted. Device data, email history, digital footprint, behavioural patterns and location signals can all be evaluated without the candidate knowing a screening layer exists. Does the email address have a credible history? Is the device linked to other applications? Do the phone number, location and claimed identity line up? Is the same person or group behind multiple supposedly unrelated submissions?
For the vast majority of applicants, this means zero added friction. The system only surfaces a flag for human review when multiple signals point in the same direction. A low-risk applicant with consistent identity signals moves through the process untouched. Higher-risk cases get closer scrutiny.
5. What role does AI play in both the problem and the solution?
AI has made every step of candidate fraud faster. A convincing resume, a tailored cover letter and a plausible interview response can be generated in seconds. It has also enabled tactics that did not exist a few years ago, like real-time face-swapping during video interviews and voice cloning that makes a proxy interviewer sound like the listed candidate.
On the detection side, AI helps identify patterns that would be impossible to spot manually across large applicant pools, such as repeated behaviours, unusual connections between submissions and inconsistencies that only become visible at volume. But AI detection only works if it has meaningful data to assess. A model running on resumes alone will miss what a recruiter misses. A model running on identity, device, behavioural and network signals catches what documents cannot.
The organizations that will handle this well are the ones that move identity verification earlier in the hiring process and connect it to real-time signals rather than static document checks. They are the ones building detection that scales with application volume without creating obstacles for the candidates who deserve a fast, fair experience.
