Imagine applying for your dream job at a top tech company, only to be judged not on your skills, but on where you went to school. A new study reveals a concerning bias in Large Language Models (LLMs) that could be making this scenario a reality. Researchers investigated how LLMs like ChatGPT, Gemini, and Claude generate "personas" for tech professionals at companies like Google, Meta, and Microsoft. They discovered that these AIs overwhelmingly assigned prestigious degrees from elite universities like Stanford, MIT, UC Berkeley, and Harvard to these hypothetical employees—a staggering 72% of the time, compared to just 8% in actual LinkedIn data. This "elite university bias" was most pronounced in ChatGPT, while Gemini fared the best. What's even more striking is this bias holds true across different career levels, from entry-level software engineers to senior VPs. This raises serious questions about the fairness of using LLMs in hiring processes, like resume screening or candidate evaluation. If these AIs are trained on data that overrepresents graduates from elite schools, they could unintentionally discriminate against equally qualified candidates from less prestigious institutions. This study doesn't just highlight the bias; it also points to ways forward. The researchers emphasize the need for more diverse training data for LLMs, as well as ongoing human oversight to identify and mitigate these biases. The ultimate goal is to ensure that AI tools promote fairness and equal opportunity, rather than reinforcing existing inequalities. The future of recruitment depends on it.