The System Never Lied. It Just Never Checked.

GCheck published its 2026 Trust in Hiring Report this month. The number at the centre of it is not subtle: 93%. That is the share of job seekers surveyed across 1,500 working adults who applied for positions in the prior 18 months, who admitted to at least one form of embellishment or misrepresentation on their application.

Read that again.

Not a suspicious minority. Not the bottom of the barrel. 93%. Across industries, experience levels, and job functions. Almost every person who has applied for a job has, by their own admission, overstated something.

The second number in the report is the one that breaks the whole picture open. Of all the discrepancies candidates introduce into the hiring process, only 26% are ever detected. Three quarters (3/4) pass through unnoticed. For every four lies told in the hiring process, three make it all the way through. They become offers. They become salaries. They become people sitting in seats they were not hired to fill.

A separate Greenhouse survey from the same period found that 62% of hiring managers now believe candidates are better at faking credentials than HR teams are at detecting them. Not equal. Better. The people whose entire job is to screen applicants believe they are already outmatched.

At its core, this is not a story about dishonest candidates. It never was.

It is a story about a system that was never designed to surface the truth. A system built on the assumption that people would present themselves accurately, that references would speak candidly, that employment dates would mean something, and that a title on a document corresponded to a real body of work. That system did not fail. It is working exactly as designed. It was designed to surface confident presentation, and confident presentation is what it surfaces.

The problem is that confident presentation is now free, programmable, and nearly indistinguishable from the real thing.

A System Built on the Honour of People With Everything to Gain

The traditional CV is a document with no external validation layer. It is written by the candidate, formatted by the candidate, and submitted by the candidate. Every word in it is chosen by the person whose interests it serves. There is no independent anchor. There is no mechanism built into the infrastructure itself that says: This claim has been checked.

References exist in theory as that external anchor. In practice, they are one of the weakest signals in hiring. Candidates choose their own references. Referees know the context. Most reference calls last less than ten minutes and produce less than one piece of useful differentiated information. Studies consistently show that positive references are almost universal, including for candidates later determined to be bad hires. The reference layer does not verify performance. It verifies that the candidate knows someone willing to say something kind.

Two Signals. Only One of Them is Real.

Here is what the data from the 2026 report has clarified in a way that can no longer be ignored: there are two kinds of candidates in any hiring pool. Those who have done the work and can prove it. And those who have not done the work but can present it. The current system treats them identically. It has no instrument to tell them apart.

The GCheck report does not describe a new problem. It describes a problem that has always existed, now measured rigorously enough that it cannot be dismissed. 93%. That figure did not appear in 2026. It existed in 2016 and in 2006. What is new is that the tools available to inflate, fabricate, and sustain the presentation have advanced by an order of magnitude.

The AI Escalation Is Not a New Problem. It Is the Old Problem Made Undeniable.

In January 2026, The Hacker News reported that the FBI had documented more than 300 US companies that unknowingly hired North Korean operatives using stolen identities and AI-generated personas, complete with fabricated work histories, deepfake video profiles, and portfolios built from nothing. Google and McKinsey, two organisations with extensive HR infrastructure, reintroduced mandatory in-person interviews as their primary countermeasure. Not better screening software. Not AI-powered detection. In-person presence, because it was the only verification mechanism they had left that AI could not yet fully simulate.

Gartner projects that 1 in 4 candidate profiles will be entirely fabricated by 2028. Not embellished. Not inflated. Fabricated from nothing. A quarter of the hiring pipeline will be fiction.

Why Existing Fixes Do Not Work

The standard response to credential fraud is more scrutiny at the point of hire: longer interviews, skills tests, and better background check vendors. Each of these solutions fails at the same structural problem. They are checks applied at the end of a process that has already been contaminated from the beginning.

Background checks are the most commonly cited safeguard. They are also the most consistently misunderstood. A background check confirms that a company with a given name existed. It confirms that a person with a given name worked there between certain dates. It does not confirm what they did. It does not confirm what they built. It does not confirm whether the outcomes listed in the achievement bullets on their CV are real, inflated, or entirely invented. Background screening was designed for identity fraud and criminal history. It was not designed to verify capability. It has never verified capability.

Skills tests at interview address a different part of the problem, but only if the skill being tested is the skill required for the job, and only if the test environment accurately reflects real work conditions. 85% of employers now use some form of skills-based hiring. But skills tests are point-in-time assessments. They do not verify the context in which skills were applied, the scale at which they were exercised, or the outcomes they produced. Passing a test and having done the work for three years at scale are categorically different things. The test cannot tell them apart.

The Cost Is Not Symmetric

The consequences of this system failure are not shared equally. The candidates who suffer most from a system that cannot distinguish genuine capability from confident presentation are not the ones doing the presenting. They are the ones who did the work, built the track record, and have no mechanism to make that track record independently verifiable.

When any candidate can claim any experience, the honest candidate has no advantage from actually having the experience. The signal has been destroyed. The people who did the work are competing on equal footing with the people who described the work. In some cases, they are losing; because the person who described the work described it better.

Verification Cannot Live Inside the Candidate's Control

The structural correction is not more vigilance. It is not better interview technique. It is not more sophisticated AI detection that will itself be defeated by more sophisticated AI generation. The structural correction is moving the verification layer outside the candidate's control entirely.

VERYFY's Capability Passport is built on a single foundational principle: proof that exists independently of the person making the claim. Verified work experience is not a candidate's self-reported work history; it is a record anchored by the organisations, managers, and artefacts connected to the actual work. Every piece of claimed experience must be tied to something that exists in the world and can be examined: a real output, a real outcome, a real person who can confirm their direct involvement. The candidate cannot write their own Capability Passport any more than they can write their own degree certificate.

Manager stamps are not references. A reference is a favour from someone who has chosen to help. A manager stamp is a formal endorsement, baked into the profile, connected to a specific role and a specific body of work, timestamped and verifiable. The manager's name is attached to the claim. The accountability structure is different. So is the incentive to be accurate.

Peer verifications close the gap that neither references nor manager stamps fully address: the lateral signal. Colleagues who worked alongside someone know things that managers often do not. They know how the work was actually done, who actually did it, and what the real contribution looked like day to day. When peers are verifying claims (colleagues whose own credibility is staked on the accuracy of their endorsement), a new layer of accountability enters the system.

Back to the 93 %

The GCheck report is not an indictment of candidates. It is a measurement of a system that gave people no reason to be honest and no consequence for not being. When the infrastructure does not verify, embellishment is rational. When embellishment is rational, it becomes universal. When it becomes universal, the signal disappears.

The only correction is an infrastructure that does not depend on the honesty of the person making the claim. Verification that is structural, not aspirational. Proof that lives in the record, not in the presentation.

The CV was always a self-report. AI just made self-reporting unlimited. The answer was never a better self-report. It was always proof. Proof as an infrastructure, not a feature.