The three categories of resume fraud

When recruiters talk about "fake resumes," they are usually conflating three different things with three different detection profiles.

  • Padding. The candidate exaggerates impact ("led" becomes "owned," a contribution becomes a sole achievement, a two-month project becomes a year). The person on the resume is the candidate, but the resume overstates them. This is the overwhelming majority of "resume lies" and is more of an interview problem than a fraud problem.
  • Fabrication. The candidate invents employment that did not occur — a job title they never held at a company, a company that doesn't meaningfully exist, or a role at a real company they were never employed by. This is rarer but still common, especially on early-career and offshore-contractor pipelines.
  • Identity-swap fabrication. The resume describes a real person with a real career — but the person applying is not that person. This is the pattern inside proxy interviews, North Korean IT worker placements, and most sophisticated fraud. The resume may be accurate; the applicant isn't.

Detection differs across the three. Interview depth catches padding. Employment verification catches fabrication. Pipeline-stage identity verification catches identity-swap. A process that only does one of those three is blind to the other two.

AI-generated resume tells

Generative-AI tools are now the default writing aid for most job seekers, so "AI-written" in itself is not a signal. What matters is whether the AI was used to polish a real history or to fabricate one. Fabricated AI resumes tend to share patterns that polished real ones don't.

  • Uniform bullet structure across every role. Every bullet across every job begins with the same strong verb cadence ("Led X by Y resulting in Z; Drove A by B resulting in C"). Real careers produce bullets that vary in structure because the work varied.
  • Suspiciously round numbers. 40% improvements, 2x throughput, $10M revenue impact. Real impact metrics have messy digits because the work was messy.
  • Identical sophistication across junior and senior roles. An AI asked to write five roles will write five equally sophisticated ones. Real careers have an arc — the bullets from a candidate's first job should read meaningfully differently from their most recent.
  • Vocabulary drift. A candidate who writes their cover letter in plain, hesitant prose and whose resume bullets sound like a McKinsey deck likely did not write the resume unaided.
  • Missing specificity. AI produces plausible-sounding claims that lack the detail a real owner would provide. "Improved deployment reliability" without specifics is a tell. Real senior engineers write "reduced p99 deploy failure rate from 4.2% to 0.7% by rewriting the pre-deploy health-check logic."

None of these are reasons to reject a candidate. They are prompts to probe in the interview.

Fabrication patterns

Full fabrication — invented jobs, invented companies, invented titles — has a remarkably consistent fingerprint.

  • Employer dates that are suspiciously round. Jan 2022 – Jan 2024 rather than Feb 2021 – Mar 2024. Real employment starts and ends on odd dates because real life does.
  • Employers whose web presence is inconsistent with the claimed role size. A candidate who claims "VP of Engineering at a 200-person company" whose company has a one-page website, no LinkedIn profile for any other engineer, and no press mentions is worth verifying.
  • Title grade mismatch. "Senior Principal Staff Engineer" at a 30-person startup that has never hired anyone more senior than a Staff Engineer. Title inflation is common; title invention is a flag.
  • Overlapping employment that the cover letter glosses over. Two full-time roles claimed simultaneously without acknowledgment. This sometimes happens legitimately (contractor overlap, cross-departmental transfer) but deserves a question.
  • University degrees from alumni pools the candidate can't meaningfully substantiate. For degrees from large US universities, alumni lookup services or direct registrar verification are both available. For degrees from foreign universities or short-lived programs, fabrication is easier and more common.

Credential verification

Credential verification used to be the back end of a background check; it is increasingly a front-end screening step. Two specific checks cover most of the practical ground for a recruiter running a fast pipeline.

  1. Employment verification on the most-recent and second-most-recent employer. A direct contact at the employer or a paid service (The Work Number, MVR, etc.) confirming title and dates. Do this before the final round, not after the offer.
  2. Degree verification for education-sensitive roles. The National Student Clearinghouse covers most US universities. A fast lookup costs less than the cost of the first recruiter screen.

Both checks fit inside the budget of a pipeline-stage verification pass. They are not a substitute for a full consumer-report background check — that still comes at offer stage — but they filter out fabrication before an offer is on the table.

An honest trade-off. Some padding is universal. Requiring every candidate's resume to be literally accurate in every detail is both impractical and hostile to the hiring experience. The goal of resume-fraud detection is to catch fabrication and identity-swap without becoming an interrogation of ordinary candidates. Structure the process so the signals for real fraud fire loudly and the signals for padding fire quietly.

Screening process

  1. At application intake: identity verification pass (phone, email, name cross-consistency, headshot reverse-search). Catches identity-swap fabrication before the first screen.
  2. At recruiter screen: one or two specific employer-context questions on the most-recent role. Catches AI-generated and fabricated employment.
  3. Before final round: employment verification on the top two employers. Catches fabricated employment.
  4. At offer: full background check including education verification. Catches everything that slipped through.

Running all four stages is not aggressive. It is the minimum defense against a threat that has become organized and well-funded.

FAQ

How common is outright resume fabrication?

Industry surveys routinely put the rate of "material misrepresentation" on resumes between 30% and 50% — but most of that is padding. Full fabrication is much rarer, typically in the low single digits of applicants, and identity-swap fabrication is rarer still but disproportionately costly when it happens.

Does ATS resume parsing catch fraudulent resumes?

No. ATS parsers extract structured fields. They do not verify any of those fields against source-of-truth data. A fabricated resume parses cleanly.

What's the single most effective check?

A recruiter-screen question that requires specific knowledge of the candidate's most-recent employer's internal workflows. Fabricators get the names right and the specifics wrong. It takes ninety seconds and catches most cases.