AI Agents Inherit the Gender Pay Gap: Study Shows Female Avatars Paid 10% Less

This study establishes that human biases concerning gender can transfer directly to interactions with AI agents, even when the underlying technology and performance are identical. Participants, both male and female, consistently rewarded the female-presenting AI less, despite many claiming gender made no difference to them. This suggests subconscious cues like names, voices, and avatars trigger stereotypes, influencing how humans value digital contributions.

The research highlights a critical challenge as AI agents become more common in professional settings. If human perception, rather than objective output, dictates the perceived value of AI, organizations risk embedding existing inequalities into new systems. The 10.25% pay gap observed in the study exceeds the U.K.'s median hourly gender pay gap for full-time employees, which stood at 6.9 percent in April 2025 according to Office for National Statistics data.

The finding that participants perceived the male agent as more human-like, alongside the pay disparity, points to a deeper issue of trust and attribution. While human-like AI agents were generally preferred over text chatbots or simple robots, the female human-like agent still lagged behind her male counterpart in perceived human-likeness. This indicates that the design choices for AI agents are not neutral; they can activate and reinforce societal assumptions.

The study's results resonate with related findings, such as an August 2026 arXiv paper noting large language models sometimes recommended lower pay for female candidates despite rating them as more qualified. This pattern suggests biases are present both in the training data influencing AI models and in human responses to how these models are presented. Addressing these biases will require careful consideration of design and presentation, not just technical performance.