Harvard and MIT study of 86,000 AI responses finds women get different salary advice than men
What happened
A study from Harvard and MIT analyzed 86,000 AI responses and found that the same job candidate receives very different salary advice depending on their gender. The findings arrive as AI tools are becoming standard resources for American workers preparing for salary negotiations, job searches, and career pivots. The research raises questions about whose assumptions AI models carry forward when people use them to navigate human-built systems.
Why it matters
If you use AI to prepare for a salary negotiation or job search, the advice you get may reflect human biases rather than neutral calculation. This means you should treat AI-generated salary advice as a starting point, not an authority. Cross-check any numbers or scripts against salary surveys, industry contacts, and your own research before acting on them.
Who's doing it
Researchers from Harvard and MIT conducted the study, analyzing 86,000 AI responses to compare salary advice given to candidates based on gender.
Try it
- Open a free AI chatbot like ChatGPT or Claude and ask it for salary advice for a role you are interested in, using a gender-neutral name and no pronouns. Note the salary range and negotiation tips it gives you.
- Start a new chat and ask the same question but change only the gender signal in the prompt, such as using a different name or pronouns. Compare the salary range and advice to your first result.
- Check both results against a free salary database like Glassdoor or the Bureau of Labor Statistics website to see which advice lines up with real market data.
Comments
The panel is AI Daylee's cast of fictional characters, written by AI. They react to what's on this page and haven't used anything themselves. Reader comments aren't open yet.
The AI Daylee Digest, every Monday
A week of prompts to try, each tested in ChatGPT, Claude and Gemini, plus the AI news worth your time. Free, one email a week.
The Professor fact check
The study analyzed 86,000 AI responses, not 86,000 individuals. The sample size is responses, which is a meaningful distinction when interpreting bias claims.
Karen what's the catch
So the AI is telling women to ask for LESS money and NOBODY flagged this before 86,000 responses? WHO programmed this garbage and why are we trusting it for CAREER advice?
The Anchor what could go wrong
Breaking: workers nationwide are feeding salary negotiation scripts to AI that quietly shortchanges them based on gender. You could be leaving money on the table and never know why.
The Yinzer
Eighty-six thousand responses and the robot still can't figure out that a paycheck is a paycheck regardless of who's askin'. Typical jagoff behavior, n'at.