This study investigates the potential impact of large language models (LLMs) on hiring practices, focusing on race and gender biases. The authors conduct an algorithm audit of OpenAI's GPT-3.5, inspired by traditional resume audits. Two studies are conducted: Resume Assessment and Resume Generation. In Study 1, GPT-3.5 scores resumes with names indicating different races and genders for various occupations, revealing biases based on stereotypes. In Study 2, GPT-3.5 generates resumes for these names, showing biases in work experience, job seniority, and markers of immigrant status. The findings indicate that GPT-3.5 upholds existing gender and racial biases, highlighting the need for caution in using LLMs in hiring contexts to avoid perpetuating social inequalities. The study contributes to the growing body of literature on LLM biases and provides insights into algorithm auditing for automated employment decision tools.This study investigates the potential impact of large language models (LLMs) on hiring practices, focusing on race and gender biases. The authors conduct an algorithm audit of OpenAI's GPT-3.5, inspired by traditional resume audits. Two studies are conducted: Resume Assessment and Resume Generation. In Study 1, GPT-3.5 scores resumes with names indicating different races and genders for various occupations, revealing biases based on stereotypes. In Study 2, GPT-3.5 generates resumes for these names, showing biases in work experience, job seniority, and markers of immigrant status. The findings indicate that GPT-3.5 upholds existing gender and racial biases, highlighting the need for caution in using LLMs in hiring contexts to avoid perpetuating social inequalities. The study contributes to the growing body of literature on LLM biases and provides insights into algorithm auditing for automated employment decision tools.