An AI warning made hiring managers rate new graduates less favorably. No one was hired.

What the study measured

Boyd-Swan and Reynolds recruited 1,750 full-time U.S. for-profit professionals with hiring responsibilities on Prolific in May–June 2026. They randomly assigned a neutral prompt, a prompt describing AI as a useful aid that can be misused, or one describing AI as a substitute for student effort and ability. Participants rated two constructed recent-graduate resumes.

The stronger substitution message reduced stated interview and salary-negotiation ratings, especially for the high-GPA resume. The milder message produced no consistent penalty. Asked what might reassure them, respondents in both AI-message groups rated in-person exams more favorably than the neutral group did.

What it cannot answer

These were hypothetical ratings from an opt-in panel. The researchers did not send applications, observe callbacks or salary offers, or verify how any student used AI. The result identifies a causal effect of the assigned message on ratings, not an AI effect on employment or earnings. It also does not show that a particular test, course or credential improves real hiring odds.

Why it matters beside the job numbers

Some early-career studies find weaker employment or hiring in exposed work; others do not find a broad unemployment jump. This experiment suggests a separate route by which an entry barrier could arise: employers may distrust what a graduate’s record shows. It has not demonstrated that this mechanism explains the observed employment gaps. A field study tracking real applications and verified assessment would narrow that gap.