AI helps people write clearer letters and tailor an application to a vacancy. But if many applicants can produce convincing text at little cost, how does a recruiter distinguish their actual abilities? Recent research examines this evaluation problem and its possible consequences for access to a first job.

What happened: a model of recruitment under pressure

Itai Ashlagi, Ramesh Johari, Jon Kleinberg and Anushka Murthy submitted a working paper on 24 September 2026 entitled “Can Labor Markets Function in the Age of AI? The Evaluation Bottleneck in Hiring”. It is based on a mathematical model, not a survey measuring actual hiring.

The authors study a scenario in which AI reduces application costs while diminishing useful information in some parts of the application. Under those conditions, companies may place greater weight on prior experience, which is easier to observe. Applicants whose qualities are initially less visible could be disadvantaged.

Why it matters: better writing does not ensure better evaluation

The result is conditional: it demonstrates a possible mechanism under specified assumptions. It proves neither a general decline in entry-level hiring nor a particular rejection rate for AI-assisted applications.

Our analysis: the issue extends beyond cover letters to the allocation of assessment time. If applications increase, a company may look for shortcuts. The risk would be to make existing experience even more decisive precisely when newcomers need their first chance to build it.

What it could change for you: make skills observable

For applicants, this research invites a distinction between presentation quality and evidence of competence. An explained project, a result placed in context or a precise discussion of one's work provides different information from polished writing. These are editorial possibilities drawn from our analysis, not methods that guarantee employment.

For employers, the authors examine multi-stage assessments that could improve selection under certain conditions. Their usefulness depends on cost and the information they provide. Adding exercises without considering their relevance could simply shift the burden to applicants.

Another perspective comes from the International Labour Organization. A brief published on 15 September and presented on 22 September studies AI adoption in Chinese companies, including changes to work organisation. Its company sample is not representative: it provides context on uses without validating the hiring model or supporting a worldwide generalisation.

Conclusion: recognising potential is the challenge

AI can make a well-presented application more accessible. It does not alone resolve the difficulty of recognising skills. For employers and applicants, the challenge will be relevant, proportionate assessments that also leave room for potential that is not yet highly visible.