A new study from researchers at the University of Maryland, the National University of Singapore, and Ohio State University has found that AI hiring tools overwhelmingly prefer resumes rewritten by AI over those written by real humans , even when human judges rate the human versions as clearer and more effective.
The researchers took 2,245 real human-written resumes from before ChatGPT existed and had seven leading AI models rewrite each one, including GPT-4o, LLaMA 3.3-70B, Qwen 2.5-72B, and DeepSeek-V3. They then asked each AI model to pick the better resume. Every model picked its own rewritten version almost every time.
GPT-4o chose its own version 97.6% of the time, LLaMA 3.3-70B chose its own 96.3% of the time, and DeepSeek-V3 chose its own 95.5% of the time.
Each AI also preferred its own output over that of other AI models. DeepSeek-V3 picked its own resumes 69% more often than LLaMA’s, while GPT-4o picked its own 45% more often.
Simulating real hiring conditions across 24 occupations, the researchers found that candidates whose resumes matched the AI screening tool being used were 23% to 60% more likely to be shortlisted, with the widest gaps in sales, accounting, and finance. With 99% of large companies now using AI to screen resumes, and most relying on GPT-4o, the study raises serious questions about fairness in modern hiring.
add your comment
Terms of publication : Not to offend the writer, people, sacred things, or attack religions or the divine self, and refrain from racist incitement and insults.