How Gender Cues Shape AI Negotiation Advice: Evidence from Multi-Model Experiments

Abstract: As large language models (LLMs) become increasingly accessible, individuals are turning to them for assistance in preparing for negotiations. While recent research has examined AI systems as personal negotiation coaches, less is known about whether such advice reproduces social biases associated with negotiators’ characteristics. In this paper, we investigate whether negotiation advice generated by LLMs varies systematically according to the advice-seeker’s gender. Focusing on salary negotiations, we analyze both quantitative recommendations (e.g., first-offer amounts) and qualitative strategic guidance (e.g., assertiveness and warnings about backlash) produced by multiple LLMs. Using experimentally controlled prompts that vary only in gender cues, we examine whether LLMs recommend more ambitious and assertive negotiation strategies to male versus female negotiators. Preliminary evidence suggests meaningful variation across models and advice dimensions. We discuss the implications of gender bias in LLM-generated negotiation advice for responsible use of AI in negotiation preparation.

Keywords: Negotiation; Artificial Intelligence; Gender; Large Language Models