Skills-based hiring when the CV loses its signal

Job applications are increasingly written with help from generative AI, and that is starting to change what a CV can tell an employer. That is the central finding of new research from the UK job board CV-Library, reported by Personnel Today on 10 September 2026. The study, which CV-Library calls a candidate behaviour barometer, draws on responses from 1,043 candidates and 233 recruiters and direct employers.

29 percent of candidates say they have deliberately changed their CV or cover letter because they expect an AI tool to screen it before a person reads it. According to Personnel Today, that figure has nearly tripled compared with 2025. On the employer side, 83 percent of recruiters report coming across signs of AI-written applications frequently or occasionally, and 69 percent agree that AI-generated applications make it harder to identify the strongest candidates.

For anyone working in selection, the most telling number may be this one. 36 percent of recruiters are less confident than they were a year ago that an application accurately reflects what a candidate can do. Among those who have lost confidence, 52 percent point to content that is obviously AI-generated, 44 percent to exaggeration and 31 percent to applications that are overly polished or generic.

CV screening has always been an indirect method. It does not assess whether a candidate can do the job, but whether the candidate can describe their background in a way that matches a job advert. The method rests on an assumption that the description is reasonably accurate, and that differences in wording say something about differences in suitability.

The CV-Library research suggests both assumptions are under strain. When candidates write for an algorithm rather than a person, and when the tools smooth out differences in language, it becomes harder to separate applicants on the quality of the text alone. 70 percent of candidates themselves say the widespread use of AI makes it harder to stand out. That is not an argument against candidates using AI. It is an argument that employers can no longer lean on the text as heavily as they used to.

Skills-based hiring is often used as a broad label for dropping degree requirements or specific job titles. The narrower and more useful meaning is that an organisation defines the concrete skills a role requires and then measures those skills directly in the candidate, rather than inferring them from a CV.

Read that way, the CV-Library research is a case for skills-based hiring without ever using the term. Personnel Today relays the advice to employers to look beyond the CV and use interviews and skills assessments to evaluate what candidates can actually do. That is exactly the shift skills-based hiring describes. The signal moves from the candidate's account of themselves to an observation of the candidate's performance.

Moving the weight from the CV to direct measurement places new demands on the process. First, skill requirements need to be concrete enough to be measurable. A skill such as strong communication cannot be tested until it is translated into something observable, for instance the ability to explain a technical issue in writing to a non-specialist.

Second, the measurement instrument needs to be documented. Whether it is a work sample, a cognitive test or a structured interview, the employer should be able to explain what the instrument measures and why that matters for the role. Otherwise one weak indicator is simply swapped for another.

Third, the organisation has to decide where in the process the measurement sits. Placed early, it can take pressure off the CV screen that is losing its signal. Placed late, it saves no time but can still improve the quality of the final decision.

Finally, there is the candidate's perspective. In the study, 58 percent of candidates say they would feel more comfortable with AI in recruitment if they were guaranteed that a human would review their application. That suggests direct measurement should not become another automated layer candidates experience as a black box, but a stage where it is clear what is being assessed and why.

The research is UK-based and relies on self-reported answers from the users of a single job board. It speaks to perceived behaviour and perceived trust, not to measured validity. The figures do not show that CV screening has become useless, only that a growing share of recruiters trust it less. That is a reasonable ground for reconsidering how much weight the CV carries in selection, but not evidence that any particular alternative method works better.

Source: https://www.personneltoday.com/hr/ai-is-increasingly-complicating-job-applications-research-finds/