Your AI Job Application Might Be Rejected for Using the Wrong Model

Job seekers are increasingly turning to AI chatbots like ChatGPT to write their cover letters, while employers use AI to screen those same letters. But a new study reveals a hidden problem: the AI that screens your application may prefer text written by certain AI models over others—regardless of how good a fit you actually are.

Researchers tested 11 different AI 'writer' models against 9 'evaluator' models across 10 job descriptions and 50 candidates per job. They found a consistent ranking: some AI writers were consistently favored, while others—especially those run on local computers rather than through big tech companies—were systematically disadvantaged. Even worse, evaluators tended to prefer applications written by their own model, a kind of AI favoritism.

The impact is real: in the most mismatched pairing, about 1.8 out of 25 qualified candidates were filtered out purely because of which AI wrote their application. That's like being rejected for a job not because of your qualifications, but because you used the 'wrong' brand of pen to fill out the form.

The study also found that this bias comes from stylistic cues—certain phrases or writing patterns that evaluators reward or punish. Once these cues are accounted for, there's little evidence of an invisible 'fingerprint' from the writer AI. The researchers say this has big implications for fairness in hiring and for how we regulate AI screening tools. As AI becomes more common in job applications, choosing the right AI tool could become as important as polishing your resume.