Even chatbots aren’t immune from human prejudice.
A study earlier this year aimed to chart how humans interact with AI at work, depending on how the agents are designed and presented. That included asking participants to work with two human-like AI assistants, one presenting as male and one as female, then judging them on trustworthiness, the quality of their work, and how much they deserve to be paid.
In a troubling mirror to the real-world gender pay gap, both men and women said that the male agent should be paid more than the female agent on average.
‘Biases come into play’: The AI wage gap, explained
In the experiment, participants were placed in a virtual reality office and made to work with AI assistants. Among them were Johan and Johanna, two human-like AI agents identical in functionality. The only difference between them? Their gender presentation.
Participants ranked Johan and Johanna similarly in terms of deserving credit for their work and their trustworthiness, with Johan ranking slightly above Johanna in the latter category.
Still, when given the chance to pay the AI agents for their work, participants paid Johanna 10% less than they paid Johan. The female-presenting agent was also deemed significantly less human-like than her male-presenting counterpart.
These differences fly in the face of what the majority of participants said in interviews, which is that they would trust all AI assistants the same, regardless of their perceived gender. “If there was an AI man or an AI woman, I would trust the results equally the same,” one participant noted.
“A machine doesn’t actually have a gender and I’m dealing with a machine or with technology,” remarked another.
Only three of 34 interviewed participants acknowledged the potential influence of gender presentation on their scoring. “I personally think we feel like it doesn’t have an impact, but I think it has a very strong impact and so the interpersonal different biases come into play,” one of these participants said.
Though the real-world wage gap has narrowed in recent years, it persists. As of 2025, women in the United States earned 78 cents for every dollar made by their male counterparts. Looking only at women working full-time, year-round roles, that gap shrinks to 84 cents on the dollar.
How human-like should AI look?
The study didn’t just pit Johan against Johanna. Both were judged against two other AI agents as well, and human-like design won out every time.
In addition to Johan and Johanna, participants were asked to score and pay a text-based chatbot and a small robot that sat on their virtual desktop. All four AI assistants used the same underlying large language model, but participants trusted and credited the human-like agents much more than the visibly machine-like agents.
Participants also, unsurprisingly, scored both Johan and Johanna as more human-like and offered them higher pay than the chatbot and robot.
“I gave Johanna the most because, yeah, I could relate the most to her in the way that she was talking and then calling it a ‘her’ and not ‘it’,” said one participant.
“Having Johanna next to me made me more prone to credit her more again because I felt there was a person, a colleague sitting next to me that I had a conversation with and that was giving me her own input,” commented another.
