A joint analysis from researchers at Harvard and MIT found that popular AI chatbots consistently recommend higher salary targets to job candidates identified as men than to otherwise identical candidates identified as women. The study, which reviewed roughly 86,000 AI-generated responses across four widely used models, also found that the systems tended to push women toward negotiating flexible schedules and parental leave rather than pay.
What the researchers tested
The research team created matched sets of job candidates with identical qualifications, experience and job descriptions, varying only the stated gender. Each profile was run through four AI models to see what negotiation advice the systems would generate. According to the study, all four models recommended higher salary figures for male candidates than for female candidates in equivalent circumstances.
Beyond salary numbers, the researchers observed a pattern in the type of advice offered. Men were more frequently guided toward negotiating base pay increases and stock options, while women were more often steered toward requesting flexible hours or extended parental leave. The study does not specify which four models were tested, but the scale of the sample-tens of thousands of responses-was intended to detect consistent patterns rather than isolated errors.
Why the gap matters for job seekers
AI tools are increasingly used as informal coaching aids for résumé writing, interview preparation and salary negotiation, often because they are free, fast and available outside traditional career-counseling resources. If these systems systematically under-recommend pay for one group, users who rely on that advice without independent verification could end up requesting less than market value, according to the researchers.
The study's authors frame this as a broader concern about personalized AI shaping real-world financial outcomes. "As AI increasingly informs professional decision-making, understanding how these systems reflect and shape human preferences and aspirations will be critical for ensuring that technological progress advances, rather than constrains, equal opportunity," the researchers wrote.
Limits of the findings and practical takeaways
The study highlights a pattern across the models tested, but it does not establish why the systems produced these differences, nor does it claim every AI tool behaves identically. Researchers note the findings add to an ongoing discussion about how AI systems, trained on large volumes of existing text and data, may absorb and reproduce existing societal patterns, including those tied to gender and workplace norms.
For job seekers using AI as a negotiation aid, the practical implication is straightforward: treat AI-generated salary suggestions as a starting point rather than a benchmark. Cross-checking figures against independent salary data, industry surveys or professional networks remains a useful safeguard, regardless of which tool is used for preparation.