Ask any hiring manager which part of a CV matters most, and the answer is almost always the same: where the person has worked, in what roles, for how long. Employment history is the spine of a candidate’s professional story. It is what salary, seniority, and the hiring decision itself are built on.
Now ask the same hiring manager how confident they are that the employment history in front of them is true. The honest answer, more often than not, is that they verified it by calling the previous employer, and that the call either confirmed almost nothing or never happened at all. The most important claim on the CV is verified by the weakest method in the screening toolkit, and that method has quietly stopped working.
This is the employment-verification blind spot, and it is widening from two directions at once. The traditional way of checking employment history has broken down, and the fraud exploiting that breakdown has industrialised.
The Most-Claimed, Least-Verified Line on the CV
The vulnerability starts with how often employment history is misrepresented. Research from First Advantage in 2026 found that 76 percent of hiring professionals reported encountering falsified employment details, making it one of the most common forms of candidate misrepresentation by a wide margin.
The misrepresentations are rarely dramatic inventions. They are quiet adjustments that survive a casual read. Dates get stretched, so a fourteen-month stint becomes “2019 to 2021.” Gaps get papered over. A junior title is rounded up to a senior one. Occasionally an employer appears on the CV that the candidate barely worked for, or never worked for at all. Each of these is plausible enough to pass unexamined, and each one inflates exactly the things a hiring decision rests on: seniority, continuity, and experience.
What makes these adjustments durable is that the verification method meant to catch them depends entirely on someone else cooperating, and increasingly, no one does.
Why the Reference Call Stopped Working
The traditional employment check is, at its core, one company asking another company a favour: please confirm that this person worked here, in this role, for this period. That favour is granted less and less often.
Former employers have no legal obligation to respond, and many have concluded that the safest answer is the smallest one. Corporate policy at a large number of organisations now prohibits sharing anything beyond confirmation of tenure, and often not even that, out of fear of defamation exposure if a candid reference goes wrong. The result is the “name, rank, and dates” reference, or no reference at all. Industry reporting puts the turnaround time for traditional phone-based employment verification at anywhere from five to fifteen business days, and a meaningful share of those requests simply go unanswered.
Here is the quiet failure mode that follows. When the previous employer does not respond, the candidate’s version of events stands by default. The hiring manager, under pressure and on a timeline, has usually decided in their head already. Three unanswered follow-ups later, the hire is approved on the strength of the CV itself. The verification did not catch the discrepancy because the verification never actually happened. It timed out. And a method that fails silently, by simply producing no answer, is worse than no method at all, because it produces the feeling of having checked without the substance of it.
The Silence Became a Market
Wherever a verification method depends on goodwill, fraud finds the seam. The collapse of the reference call created a market, and that market is now well supplied.
Paid reference-for-hire services exist openly, offering to play the part of a former manager for a fee. AI has made the fakes cheaper and more convincing still: a fabricated referee can now come complete with a plausible LinkedIn profile and a real-looking email address. Screening practitioners have learned to treat an insistence on email-only contact as a red flag, precisely because a made-up reference prefers to avoid a real-time conversation where they might slip. But the underlying problem is structural. A verification system built on “we will believe the person the candidate tells us to call” is trivially gamed by a candidate willing to choose who answers.
The favour-based model was always vulnerable to this. What changed is that fabricating a convincing referee went from difficult and risky to cheap and scalable.
Now the Phone Call Itself Can’t Be Trusted
Even that assumes the worst case is a real human being coached to lie. The newer problem is that you can no longer be sure there is a real human at all.
AI voice cloning has crossed the threshold from novelty to commodity. A usable clone can now be built from as little as a few seconds of audio, and the tools are cheap, fast, and require no skill. Voice-phishing attacks surged sharply through 2025 as a direct result, and voice-biometric defences fail to detect synthetic voices in a meaningful fraction of cases. The canonical warning is the 2024 incident at the engineering firm Arup, where an employee authorised transfers totalling roughly 25 million dollars after a video call in which every other participant, including a senior executive, was an AI-generated fake.
Translate that into the reference context. A live call to a “former manager” used to carry an implicit guarantee: at least you knew you were talking to a real person in real time. That guarantee is gone. The voice can be synthetic, the warmth can be generated, and the specifics can be scripted by a language model that read the candidate’s CV. Security researchers call the broader effect “truth decay,” the point at which no digital interaction can be trusted at face value. The reference call sits squarely inside that decay.
The Scale of the Problem
This breakdown is not arriving in a quiet period. It is arriving as candidate fraud industrialises.
The screening firm Checkr has reported that 23 percent of companies have already encountered identity fraud among new hires. Gartner has projected that by 2028, as many as one in four candidate profiles worldwide could be fake. The pressures behind those numbers, AI-generated resumes, synthetic identities, organised schemes running candidates through multiple remote roles at once, are the same pressures that make a fabricated employment history easy to assemble and a fabricated referee easy to supply.
So the two curves cross at the worst possible angle. The method for verifying employment history is getting weaker exactly as the incentive and the tooling to falsify it are getting stronger. An organisation relying on the reference call is not holding steady. It is falling behind on both axes at once.
From Goodwill to Records
The instinct, faced with all this, is to try to fix the reference call: make more attempts, ask sharper questions, train interviewers to spot a scripted referee. None of that is wrong, and some of it helps at the margins. But it misunderstands the nature of the failure. The reference call is not broken because people are doing it badly. It is broken because it depends on a favour that is no longer granted and on an assumption, that the voice is real, that no longer holds.
The way out is not a better favour. It is a different source of truth. Instead of asking a former employer to vouch for a candidate, verify against the records that employer already created at the time, contemporaneously, and reported to a neutral third party. Payroll data. Statutory employment records. The contributions an employer made to a government scheme month by month while the person actually worked there. These records do not depend on anyone picking up the phone, they cannot be talked into saying something generous, and they cannot be cloned.
That shift, from verifying by goodwill to verifying by record, is the subject of the companion playbook, and it is the single most consequential change an organisation can make to close this particular blind spot. The reference call had a good run. The records were always the better witness.
