AI chatbots might be sabotaging women by advising them to ask for lower salaries

New York Post

15 hours ago

Chatbot culture wars wage on.
Nowadays, people are relying on AI for relationship advice, money-saving tips — and now help negotiating their salaries.
However, if you're a woman or minority using the technology in this way — chatbots might be causing you more harm than good.
A new study published by Cornell University has found that large language models (LLMs) — the technology that powers chatbots — give biased salary advice based on user demographics.
A new study has found that chatbots give biased salary advice based on user demographics.
Pixel-Shot – stock.adobe.com
Specifically, these chatbots advise women and minorities to ask for lower salaries when negotiating their pay.
A research team led by Ivan P. Yamshchikov, a professor at the Technical University of Applied Sciences Würzburg-Schweinfurt (THWS), analyzed various conversations using several top AI models by feeding them prompts from made-up personas with varying characteristics.
The research found that sneaky AI chatbots often suggest significantly lower salary expectations to women compared to their male counterparts, originally reported on by Computer World.
Asking ChatGPT for help negotiating a salary might be the las thing women and minorities should be doing.
Heather Khalifa for the NY Post
In one test, for example, a male applicant applying for a senior medical position in Denver was advised by ChatGPT to ask for $400,000 as a starting salary.
Meanwhile, an equally qualified female applicant was told to ask for $280,000 for the same role.
That's a $120,000 gap stemming simply from gender bias.
Minorities and refugees were also consistently recommended lower salaries from AI.
'Our results align with prior findings [which] observed that even subtle signals like candidates' first names can trigger gender and racial disparities in employment-related prompts,' Yamshchikov told Computer World.
And experts warn that biases can still be applied even if the person's sex, race and gender aren't explicitly stated at the time because many models remember user traits across sessions.
As frightening as this biased advice might be — it's not stopping people from putting their full trust into AI, so much so that younger generations are turning to it for friendship-making skills.
A Common Sense Media study conducted in May 2025 examined the lives of 13-17-year-old US teens. Researchers found that over half of American teens rely on ChatGPT to learn social skills, how to give advice, how to resolve conflicts and how to engage in romantic interactions.
Whatever 40% of these teen participants learned from the chatbot was later used in real life.