A fabricated employment history is rarely caught by reading a CV carefully. It is caught by testing the CV against itself. Invented careers are internally inconsistent in predictable ways: arithmetic that does not reconcile, seniority with no trail behind it, responsibility described without the specifics that doing the work would have produced.
What follows is a method rather than a list of red flags, because isolated flags produce false accusations and the method produces evidence.
Start With Arithmetic, Not Judgement
The cheapest tests are numerical and require no interpretation.
Do the Dates Reconcile?
Total the tenures. Check for overlapping full-time roles, gaps presented as continuous employment, and tenure that predates a stated qualification the role required. Month-level precision matters: histories given only in years are frequently hiding a reconciliation problem.
Is the Progression Physically Possible?
Map claimed seniority against elapsed time. A head of function with six years total experience is possible in a small organisation and improbable in a large one. The question is not whether it could happen but whether the earlier roles that would produce that capability are present.
Does the Scope Match the Employer?
A claim to have managed a thirty person team at an organisation with forty staff is checkable in minutes. So is a claim to have owned a budget larger than the employer's plausible turnover.
Then Test Specificity
This is the core of the method. Fabrication is fluent about responsibility and silent about detail, because responsibility can be asserted and detail has to be remembered.
Read each role description and ask what a person who actually did this would inevitably know. Then check whether any of it is present.
- Named systems, tools, frameworks or suppliers rather than categories.
- Scale expressed in numbers: volumes, headcount, budget, sites, transactions, caseload.
- A constraint they worked within: a deadline, a budget cut, a regulatory requirement, a legacy system.
- A decision they personally made, as opposed to a process they were part of.
- An outcome with a direction and a magnitude, not simply an improvement.
A genuine history contains these unevenly, because real careers have detailed patches and vague ones. A fabricated history is uniformly smooth and uniformly unspecific. Evenness is itself a signal.
The Claimed Versus Evidenced Distinction
Formalise the reading. For every essential requirement, record one of three outcomes: the CV evidences it with corroborable specifics, the CV claims it without evidence, or the CV does not address it.
Two things follow. Candidates cannot be compared on vocabulary, which is what unstructured sifting effectively does. And the interviewer receives a list of exactly what needs probing, which is the most valuable output of a sift and the one most often discarded.
At volume this has to be systematic rather than intuitive. The failure mode of manual sifting is not that reviewers cannot spot fabrication, it is that attention degrades across a large pile and the last applications receive a scan rather than an assessment.
Signals Worth Checking Externally
The Employer Existed as Described
Companies House gives incorporation and dissolution dates in minutes. A role claimed at a company two years before it was incorporated is conclusive rather than suggestive.
Registration Is Current
Where a role depends on registration, check the register rather than a certificate. Nursing and midwifery, medical, allied health, solicitors and the engineering institutions all publish searchable registers. This is the single highest-value check available and it takes minutes.
Referee Details Are Independent
Referee email domains that do not belong to the employer named, or phone numbers shared between supposedly unrelated referees, are among the strongest signals available. Establish contact details independently rather than using what the candidate supplied.
Interviewing Against a Suspected Gap
The technique is depth on one point rather than breadth across many. Choose a single claimed achievement and go three questions further into it: what the alternative option was, who disagreed, what it cost, what you would do differently.
Someone who did the work can keep going, because the detail exists to recall. Someone reciting a constructed narrative runs out, and the failure is characteristic: the answers become more general as the questions become more specific, which is the opposite of what happens with genuine recall.
Ask about failure too. Fabricated histories are uniformly successful. Real work contains an overrun, a compromise or something that did not work.
Where This Goes Wrong
Three mistakes turn a reasonable process into an unfair one.
Treating a career break as a discrepancy. Illness, caring responsibilities, redundancy and study produce gaps. A gap is a question, not a finding, and treating it as evidence of dishonesty discriminates in practice against women, disabled people and carers.
Penalising presentation. Non-native English writers, people from less advantaged educational backgrounds and neurodivergent candidates often present genuine experience clumsily. Fabricated CVs usually read better than real ones. Assess evidence, not polish.
Varying scrutiny by person rather than by role. Escalating verification because of a name or an accent is unlawful under the Equality Act 2010 and also ineffective. Apply the same depth to every application for a given role, and record that you did.
Frequently Asked Questions
What Should We Do If We Suspect Fabrication?
Do not accuse. Put the specific inconsistency to the candidate and invite an explanation, in writing where possible. Many apparent discrepancies have mundane causes. Record the question, the answer and the decision.
Can We Reject Purely on a Discrepancy?
You can reject on a documented, role-relevant assessment. Take advice before treating a discrepancy as dishonesty, particularly where the candidate has offered an explanation, and keep the record of what was assessed.
Do Employment Gaps Indicate a Problem?
No. Gaps are extremely common and have ordinary explanations. What matters is an unexplained gap presented as continuous employment, which is a reconciliation failure rather than a gap.
Are AI Detectors Useful Here?
No. They are unreliable, biased against non-native writers, and answer a different question. Whether a CV was drafted with assistance says nothing about whether the history is real.
How Long Should We Keep the Assessment Record?
Long enough to defend the decision, which in practice means retaining it alongside the application under your documented retention schedule. Keep the rationale, not merely the outcome.
The Method in One Paragraph
Reconcile the arithmetic. Test whether each claim carries the specifics that doing the work would have produced. Classify every requirement as evidenced, claimed or absent. Verify externally wherever a register exists. Establish referee contact details independently. Probe depth on one point rather than breadth across several. Apply all of it identically to every candidate and write down that you did.
That process catches fabrication, and on honest applications it produces a better shortlist than reading ever did.
This article is general information, not legal advice. Take specialist employment law advice before acting on suspected dishonesty.
Sources
Nursing and Midwifery Council. Check the Register.
https://www.nmc.org.uk/registration/search-the-register/
Health and Care Professions Council. Check the Register.
https://www.hcpc-uk.org/check-the-register/
Solicitors Regulation Authority. Solicitors Register.
https://www.sra.org.uk/consumers/register/
Engineering Council. UK-SPEC.
https://www.engc.org.uk/ukspec
The Stationery Office. Equality Act 2010.
https://www.legislation.gov.uk/ukpga/2010/15/contents
GOV.UK. Equality Act 2010: Guidance.
https://www.gov.uk/guidance/equality-act-2010-guidance
Information Commissioner's Office. Employment Practices: Recruitment and Selection.
https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/employment/recruitment-and-selection/
InsightCircle
