Nine in 10 CVs now contain AI-generated inconsistencies — and recruiters are fighting back with AI of their own
The News
A new analysis from hiring platform Cadient examined 3,000 CVs in its database and surfaced troubling patterns: nine out of 10 salaried CVs contained logical inconsistencies such as overlapping dates and impossible promotions. Three out of four applications submitted to what Cadient considers top employers had AI-generated content. Perhaps most striking: two in five applications would be auto-rejected if a recruiter verified the basics listed in the CV.
Bill Mastin, CEO of Cadient, described the current recruitment landscape as "a battle between AI and AI." Candidates are feeding job descriptions and their own CVs into large language models, producing perfectly tailored applications that all read like unicorns. "From a recruiter's perspective, they get 1,000 CVs, and they have 1,000 unicorns," Mastin said. Recruiters are now deploying their own AI tools to distinguish genuine qualifications from AI-generated flimflam.
Katherine Loranger, CHRO at Safeguard Global, emphasised that screening tools alone cannot solve the problem. The interview remains the critical gate: recruiters need to ask not just what candidates did, but how they did it — probing for the human judgement and situational detail that AI-generated responses cannot convincingly fabricate. "We want people to come across as human in our interviews," Loranger said, "and we want people to not be perfect and show us who they really are."
Why It Matters
For TA leaders, this is not a future problem — it is the current operating environment. The Cadient data confirms what many recruiters have sensed for months: the signal-to-noise ratio in inbound applications has collapsed. When every application looks like a perfect match on paper, traditional screening methods break down. The cost is measured not just in wasted recruiter time but in bad hires who slip through — RJ Frasca of Shield Screening noted that hiring the wrong person can push work onto skilled employees, driving them to leave out of frustration.
The strategic question is whether your screening process was designed for an era when applicants used word processors and spellcheckers — not large language models that can rewrite entire CVs to match each job description. Organisations that treat this as a technology problem (buy another AI screening tool) will miss the point: the solution is partly technological, but it is equally about redesigning interview protocols, training hiring managers to probe for authentic experience, and setting clear expectations with candidates about acceptable AI use during the application process.
Taking Action
All action is situated in your corporate context. These are starting points, not prescriptions — choose what fits.
Connect
Who to talk with in your enterprise
Your head of talent acquisition operations — review whether your current screening stack can detect AI-generated inconsistencies at scale, and ask your ATS provider what signals they surface when a CV has been heavily modified by a language model.
Your learning and development lead — design an interviewer refresher that shifts focus from "what did you achieve?" to "walk me through how you did it, step by step," with concrete examples of questions that surface authentic experience rather than rehearsed AI output.
Inspect
What to take a closer look at
Pull a random sample of 100 CVs from your last quarter's applications and manually check for the three patterns Cadient flagged: overlapping employment dates, implausibly rapid promotions, and identical phrasing across different applicants. Compare your findings against whatever AI screening signals your current platform provides. Specific tool to look at: Your ATS's duplicate detection and content similarity reports — most major platforms (Workday, SuccessFactors, iCIMS) have these but they are often configured with thresholds too loose to catch AI-generated near-duplicates.
Reflect
What to think about more deeply
The most interesting question isn't "how do we catch AI-generated CVs?" — it's "what does an authentic application look like in 2026, and are we asking candidates to pretend they didn't use the same tools we expect our own teams to use daily?" If the candidate experience requires people to hide the tools that make them productive, what does that say about the employment relationship we are building from day one?