CVs are the staple of recruitment and CV screening has been our principal sifting and shortlisting method for decades. We’ve asked candidates to evidence their experience, qualifications and skills on a page or two, and we’ve screened, ranked and managed entire pipelines on the strength of it.

But they’ve stopped working.

And the uncomfortable truth is that the problem didn’t start with AI. AI simply finished the job.

AI has broken CV screening

Generative AI now can now match the experience and skills listed in a CV to role requirements so well that across the top half of any large application pool, we’re hearing reports from clients that they often cannot find a meaningful difference between one CV and the next. The documents are polished, keyword-aligned and structurally identical and the genuine signal of candidate potential is buried underneath.

Application volumes compound the problem. AI has reduced the effort required to apply, and employers are seeing application numbers increase as a result. According to ISE, employers receive 140 applications per graduate vacancy on average, a year-on-year increase of 14%. Roles in FMCG and tourism went up to 290 applications on average. The result is more CVs to sift with less differentiation between them.

Then there’s authenticity. Determining whether any individual CV has been AI-augmented to the point of misrepresentation is challenging and policing it with AI word-matching tools means using AI CV screening software to sift AI-generated documents. That isn’t assessment; it’s an arms race with no winner.

One way or another, we are going to be forced to change how we sift CVs and shortlist candidates.

CV screening was never predictive, or fair

Here is the part that should make the change easier: selection research has been telling us for decades that CVs were never a good sifting tool in the first place.

Meta-analyses of selection methods consistently place the things CVs showcase, years of experience, education, prior job titles, among the weakest predictors of actual job performance, far behind structured interviews, work samples, and ability-based assessment.

Worse, CVs actively invite bias. A landmark multi-national field study, reported by Oxford’s Centre on Migration, Policy and Society, found that identical CVs sent under different names produced dramatically different outcomes. Applications with British-sounding names (think James or Hannah) received significantly more positive responses than applications with names associated with non-white minorities (like Abdul or Yasmeen). Plus, candidates from ethnic minority backgrounds had to submit 60% more applications to receive the same number of positive callbacks as white British applicants. Most damning of all: this level of discrimination has remained essentially unchanged in the UK for half a century.

As our Head of Assessment Design, Amanda Callen, puts it: “Employers don’t intend to make shortlisting judgements based on demographics, but CV sifting has always been more of an art than a science, and we know that bias and unintentional discrimination occurs most easily when the source being evaluated is variable, and particularly when the basis for evaluation is unstructured and holistic, as it usually is when CVs are reviewed. Recruiters typically have to rely only on their own interpretations, inferences and judgements, and in these conditions, it’s a real challenge to be consistent, objective and focused only on job-relevant priorities.”

There was a brief, hopeful theory that wide availability of Gen AI might level this playing field by helping every candidate to produce a polished application. In practice, it has been the final nail in the coffin. AI hasn’t fixed the CV’s flaws, it has standardised them. The majority of CVs now carry the same polish, so what’s left is a document with less to distinguish candidates than ever. Meanwhile the details that trigger bias, names, schools, career gaps are all still on the page.

Alternatives to CV screening: building an AI-resilient sift

To be fair to the CV, it has survived this long for real reasons. It’s free, universal and portable. Candidates know how to produce one, and every system in the recruitment tech stack is built to ingest one. CVs also retain genuine usefulness later in the process, as context for interviews, for verifying qualifications and employment history, and for screening obligations in regulated roles.

But none of that rescues it where it does the most damage: the initial sift. There, the CV is unstructured, saturated with bias-creating details, and impossible to score consistently against the requirements of the role, it’s unlikely two candidates will present the same work history.

The good news is that the alternatives don’t just patch the AI problem, they fix weaknesses the CV has had all along. One proven route is the enhanced, bespoke application process: structured application questions built on deep insight into the skills, motivations and values that genuinely align someone with the role and organisation. With thoughtful scoring protocols, these can be automated within existing ATS platforms, delivering scalable, consistent, role-relevant evaluation at scale without a single keyword-matching shortcut. To be clear, this is not automated CV screening rebadged: what’s being scored is structured, role-specific responses, not free-form documents ranked by keyword.

“It’s not that CVs don’t contain any relevant information, it’s that they also contain so much irrelevant information, and keeping focused on what really matters while ignoring what doesn’t is a real challenge for recruiters.  A structured, bespoke application form allows only job-related predictive criteria to be evaluated and eliminates the risk of non-relevant information interfering with the validity of the evaluation being made.” 

Amanda Callen CPsychol AFBPsS FRSA, Head of Assessment Design, PeopleScout

Done well, this approach is:

• More predictive: Every candidate answers the same role-relevant questions, scored against the same criteria
• Fairer: It does not include biographical details that trigger bias, widening access for under-represented and socially disadvantaged groups
• AI-resilient: Responses grounded in personal motivation and judgement are far harder to outsource convincingly than a CV
• Better for candidates: Transparent, engaging, and capable of returning fast, personalised feedback to every applicant, not just the successful ones

Proof it works: sifting without the CV

This isn’t just theoretical. We did exactly that when TMP Worldwide – the multi-award-winning recruitment advertising agency – launched its first-ever graduate programme. Assessment Works™ by PeopleScout designed the selection process from scratch: no historical data, unknown application volumes, and an explicit requirement to be Gen AI-aware. The early sift didn’t touch a CV. Instead, candidates answered autoscored situational and preference questions, built from a values-led assessment framework and designed specifically to limit Gen AI disruption, evaluating genuine responses and different thinking styles rather than polished documents. The highest scorers progressed to asynchronous video interviews, then to an immersive assessment centre built around realistic work samples at TMPW’s London headquarters.

We sourced graduate hires who were both exceptionally talented and genuinely motivated to succeed at TMPW without dropping the efficiency of automated sifting. The results: 100% assessment centre attendance, 100% offer acceptance, and positive candidate feedback on the experience, all through bespoke assessments with no cost-per-use which kept costs down. This kind of structured sifting doesn’t just survive contact with an AI-enabled candidate market, it outperforms the old CV screening methods.

Read the full case study here.

Letting go of the CV screening comfort blanket

Nothing in recruitment is perfect, and no alternative to the CV will be a silver bullet. Building the sift and shortlisting tools that work for your context will take exploration, trialling, and tolerance. But on the evidence, decades of it, one thing is certain: we can find something better than the CV, because almost anything structured is.

AI in the hands of candidates isn’t recruitment’s enemy, no matter how it feels when we’re pressing on with outdated tools.

This is the catalyst we’ve needed for years, the push to build more predictive, fairer and AI-resilient ways of finding the people who will build our future organisations.

Ready to rethink your sift?

Assessment Works™ by PeopleScout helps organisations design assessment and sifting processes that are predictive, inclusive, and AI-resilient, from structured application design and scoring protocols to full selection process transformation. If CV screening is leaving you with identical shortlists, let’s talk.