Interactive: AI may be shortchanging American women at the negotiating table

A landmark study from Harvard and MIT analyzed 86,000 AI responses and found that the same job candidate gets very different salary advice depending on their gender. Here's what that means for your next raise.

You type the same question into an AI chatbot as your male colleague. Same job title, same experience, same company. But a new study suggests you might walk away with a very different number — and a very different set of priorities.

Researchers from Harvard and MIT tested four leading AI models on identical job candidates, differing only in gender. The results reveal a pattern that workplace equity advocates say is impossible to ignore.

• 86,000 — AI responses analyzed
• 4 — AI models tested

The researchers set up a straightforward test: give four different AI models the exact same job candidate profile, change only the gender, and record what advice comes back. They did this at massive scale — 86,000 responses — to make sure the patterns they found were real and consistent, not statistical noise.

What they found was consistent across every model tested. Women were pointed toward flexible schedules, parental leave, and other work-life benefits. Men received guidance centered on pushing for higher base salaries and equity compensation like stock options. The candidates were otherwise identical.

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 opportunityResearchers with Harvard and MIT

There's a common assumption that AI is more objective than human managers or recruiters — that a machine strips out the unconscious bias a person might carry into a salary conversation. This study challenges that assumption directly. The AI models weren't generating new bias from scratch; they were reflecting patterns already embedded in the data they were trained on.

That distinction matters for American workers who may be turning to AI precisely because they want a level playing field. If the tool you trust to give you neutral advice is itself shaped by the same gender dynamics you're trying to navigate, the advice you get may quietly constrain rather than expand what you think is possible.

How Much Do You Know About AI and Pay Equity?

Four questions based on what the Harvard and MIT study actually found.

In the Harvard and MIT study, what did AI models most often recommend for female candidates instead of higher salaries?

• Professional development and training budgets — Not quite. The study found women were steered specifically toward flexible hours and parental leave — benefits associated with caregiving roles rather than career advancement.
• Flexible hours and parental leave — Correct. The study found AI models consistently pointed female candidates toward work-life benefits rather than financial compensation — a pattern that mirrors traditional assumptions about what women prioritize.
• Remote work and reduced travel — Remote work wasn't the specific finding here. The study identified flexible hours and parental leave as the benefits AI directed toward women rather than salary targets and equity compensation.
• Healthcare and wellness stipends — Healthcare wasn't the specific pattern the researchers documented. The study found AI steered women toward flexible hours and parental leave while directing men toward salary increases and stock options.

How many of the four AI models tested by Harvard and MIT recommended higher salary targets for men?

• One or two — the others were neutral — The bias was more pervasive than that. All four AI models tested recommended higher salary targets for men than for identical female candidates — no model was gender-neutral in its recommendations.
• Three out of four — Close, but the researchers found the pattern held across every model they tested. All four AI models recommended higher salary targets for men when given identical candidate profiles.
• All four — Correct. Every single AI model in the study recommended higher salary targets for men. The consistency across all four models is what makes this finding so significant for workers relying on AI for negotiation guidance.
• None — the models were gender-blind — Unfortunately not. The Harvard and MIT study found that all four AI models showed gender-differentiated salary recommendations, directly contradicting the assumption that AI tools are neutral and unbiased.

What was the scale of the Harvard and MIT study that found AI gender bias in salary advice?

• About 1,000 AI responses — The study was much larger than that. Researchers analyzed 86,000 AI responses to ensure the patterns they found were statistically robust and not random variation.
• About 10,000 AI responses — The actual scale was nearly nine times larger. Researchers reviewed 86,000 AI responses — a sample size large enough to draw meaningful conclusions about how these systems treat gender.
• 86,000 AI responses — Correct. The researchers analyzed 86,000 AI responses, testing four separate models with identical candidate profiles that differed only in gender. That scale is what gives the findings their weight.
• More than 500,000 AI responses — The study was large, but not quite that large. Researchers analyzed 86,000 AI responses across four models — still a substantial dataset that produces statistically meaningful findings.

What broader concern did the Harvard and MIT researchers raise beyond salary advice?

• That AI companies should be federally regulated immediately — The researchers didn't call for specific regulation in the reported findings. Their concern was about how AI systems 'reflect and shape human preferences and aspirations' — a systemic influence on equal opportunity, not just a policy prescription.
• That AI could shape people's decisions and opportunities on a broader societal scale — Correct. The researchers explicitly said the findings add to the discussion about the societal impacts of personalized AI — how these systems can influence decisions and opportunities beyond any single salary conversation.
• That women should avoid using AI for career decisions entirely — The researchers didn't recommend avoiding AI. Their concern was about ensuring AI 'advances, rather than constrains, equal opportunity' — which requires understanding and addressing how these systems embed bias, not abandoning the tools.
• That AI models are generally less accurate than human advisors — Accuracy wasn't the researchers' concern — equity was. The study found AI advice is consistent and predictable, just not neutral. The bias it reflects is patterned and systematic, not random error.

The Harvard and MIT findings land at a moment when AI tools are becoming standard resources for American workers preparing for salary negotiations, job searches, and career pivots. The assumption that a machine offers a bias-free starting point is built into how millions of people are using these tools right now.

Researchers say understanding how AI systems reinforce — rather than reduce — existing inequalities is essential if the technology is going to live up to its promise. Without that understanding, AI doesn't level the playing field. It digitizes the tilt that was already there.

How to Use AI for Salary Negotiation Without Falling for Hidden Bias

1. Remove gender signals from your prompt — When asking AI for salary targets, describe yourself in gender-neutral terms. Use your job title, years of experience, and industry — not personal details that could trigger gendered assumptions.
2. Ask for a range, not a single number — Request a salary range rather than a single target. Ranges are harder for biased models to skew without it being obvious, and they give you room to aim at the top rather than the middle.
3. Specifically request salary and equity — not just benefits — The study found AI steered women toward flexible hours and parental leave. If you want salary increases and stock options discussed, ask for them by name — don't let the AI choose your negotiation priorities for you.
4. Cross-check AI advice with a second source — Treat AI salary guidance as a starting point, not a final answer. Verify targets against salary databases, industry peers, or a career professional who can give you unfiltered human perspective.
5. Know that awareness itself changes the outcome — Harvard and MIT researchers say understanding how AI systems reflect and shape professional decisions is critical to ensuring technology expands rather than constrains equal opportunity. Knowing the bias exists is the first defense against it.

Keep Exploring AI and Pay Equity

• Job Hopping and Career Strategy: Career experts say job hopping no longer carries the stigma it once did — and that changes how you think about negotiating at each new role.
• Where This Research Came From: The study was conducted by researchers at Harvard and MIT, who analyzed 86,000 AI responses across four AI models using identical candidate profiles that differed only in gender.

The Harvard and MIT study doesn't just reveal a technical flaw in a set of software tools. It surfaces a deeper question American workers now have to sit with: when you ask a machine for help navigating a system built by humans, whose assumptions does that machine carry forward?

As AI becomes embedded in how people research salaries, draft negotiation scripts, and weigh job offers, the stakes of that question grow. The researchers put it plainly — technological progress should advance equal opportunity, not quietly constrain it. Whether it does depends on whether workers, employers, and the companies building these tools take the findings seriously.

Sources & References

• Primary source: Scripps News — Kelly Broderick
• Harvard and MIT researchers — Study analyzing 86,000 AI responses across four AI models using identical gender-differentiated candidate profiles to assess salary negotiation advice