AI vs. AI: Job Seekers Who Use the Same Chatbot as the Recruiter Get the Edge

The job market has a new, invisible bias: artificial intelligence favors resumes written by other AI. Imagine two candidates with nearly identical profiles. Same experience, same skills, same career path. The only difference? One wrote their resume themselves. The other polished it with a generative AI tool like ChatGPT. Now imagine the first screening filter isn't a person, but another AI. Who moves forward? According to new research from the Future for Work Institute, the answer is often the AI-assisted candidate. Algorithms systematically prefer text that aligns with their own linguistic patterns. In Galicia, Spain, there’s a saying: “xente con xente” (people with people), meaning birds of a feather flock together. AI does the same. It looks for its own kind—and right now, it recognizes style, not talent. A study from the University of California tested this directly. Researchers took 2,245 human-written resumes and rewrote them using models like GPT-4o, LLaMA, and DeepSeek. When algorithmic evaluators scored the resumes, they preferred the machine-written versions between 67% and 82% of the time. The irony deepens. A separate study from Princeton’s Center for Digital Humanities found that AI models actually prefer human-written text about half the time—similar to human preference. But here’s the catch: when told “this was written by a person,” the models showed a 34-point bias toward human authorship. That’s 2.5 times stronger than the bias shown by human evaluators (13.7 points). The machines aren’t even consistent with themselves. This creates a new divide in the labor market. We already know about inequalities tied to education, experience, and networking. Now add another: knowing which AI tool to use to make your application attractive to another AI. **Students are catching on** The pressure is showing up in classrooms. This spring, some graduates in San Francisco booed AI at their commencement ceremonies—a sign of real anxiety. But most students aren’t rejecting the technology. They’re demanding to be taught how to use it properly. Data from the Kogod School of Business at American University shows regular AI use among business students jumped from 6.2% to 29% in three years. Meanwhile, the share of job interviews including AI-related questions has surged from 11.6% to 42.6% in the same period. The anxiety is also visible in employment numbers. The European Central Bank reports youth unemployment in the eurozone rose to 15.1% in the first quarter of 2026, even as overall unemployment fell to 6.3%. The ECB cautions against blaming AI directly, but notes one uncomfortable exception: youth employment in the tech sector has dropped 18.6% since 2023—far more than a simple economic slowdown would explain. **New rules, same game** The European Union is stepping in. The AI Act, which took effect last week, requires companies to ensure transparency and bias detection when AI is used to screen resumes, evaluate skills, or rank candidates. The law doesn’t ban these tools, but it mandates human oversight. Candidates now have the right to know if an AI is involved in their application process and what role it plays. But oversight isn’t the same as fairness. The rules of the game—which resume gets through, which model judges it, what “sounds good” to a machine—are still set by the companies selling the technology and the businesses buying it to cut costs. The candidate who doesn’t know which AI the company uses, the student without access to quality training, and the worker whose employer won’t invest in explaining the system that evaluates them: these are the ones left behind. Kogod’s interim dean, Casey Evans, told Axios that when ChatGPT first appeared, the priority was protecting exams from cheating. Now, she says, they’ve accepted that students will use it. Teachers are completely rethinking their role in the classroom. The message is clear: AI is becoming the intermediary for decisions that shape our lives. Learning to live with it critically isn’t optional anymore. The students know it. The recruiters know it. The question is whether the systems we build will ever learn to see talent beyond the style of the text.