Nvidia AI chief reveals how to get past automated hiring systems

Applicants using resumes generated by the same AI model as their recruiter's screening system are up to 60% more likely to get shortlisted, according to research cited by Nvidia Chief Software Architect Jonathan Ross at the Sohn Investment Conference 2026.

The finding exposes a troubling bias baked into algorithms now widely used to filter millions of job applications.

Ross, who previously worked on Google's TPU chip, framed the problem starkly: "AI likes to use AI." Recruiters increasingly deploy large language models to automate resume review, but these systems consistently favor applications written by their own underlying models, a phenomenon researchers call "self-preferencing."

A late-2025 study titled "AI Self-preferencing in Algorithmic Hiring" tested over 2,200 resumes across 24 job categories to quantify the effect. The results were stark: candidates whose resumes were generated by Claude or GPT-4 faced different approval rates depending on which model screened their application.

Human-written resumes with identical qualifications were rejected far more often, regardless of actual merit.

Each AI model was trained on specific datasets and optimized to recognize patterns in text generated by systems like itself. When presented with resumes matching those patterns, the evaluating algorithm simply finds them more legible and compelling.

Ross's advice is pragmatic but unsettling: applicants need to game the system by creating multiple AI-generated versions. "Build one résumé with Claude or Opus 4.7 and one with ChatGPT, and you'll have the highest probability of being selected," he said.