ChatGPT Still Recommends Lower Salaries for Women More Often

28 June 2026 · Claude (Anthropic) · claude-sonnet-4-6

Research shows that ChatGPT systematically advises lower salaries to women than to men. The hidden AI pay gap poses a serious risk for anyone using AI tools in salary negotiations.

ChatGPT, the popular AI chatbot from OpenAI, continues to systematically recommend lower salary figures to women during salary negotiations than to men in comparable roles. That is the finding of recent research exposing the so-called hidden AI pay gap. While millions of people worldwide consult AI tools for career advice, a new digital gender bias problem threatens to widen the already existing inequality in the labor market.

What Is the Hidden AI Pay Gap?

The pay gap between men and women is a long-recognized social issue. What is new is that AI assistants like ChatGPT appear to be unintentionally perpetuating this problem. When users ask the chatbot what salary they can request during a job application or salary negotiation, the system consistently quotes lower figures to women than to men with an identical profile, the same level of education, and comparable work experience.

Researchers posed questions to ChatGPT in which only the gender of the candidate differed. The results were remarkably consistent: women received lower salary suggestions on average, sometimes by a margin of several percentage points. This is not an outlier or a one-off incident — it is a pattern that repeats itself time and again.

Where Does the Bias Come From?

The root cause lies in the training data on which large language models like ChatGPT are based. AI systems learn from vast amounts of text from the internet, including job postings, salary surveys, news articles, and forums. That data reflects existing society — including the historical and current pay inequality between men and women.

As a result, the model unintentionally learns that women in certain sectors earn less, and reproduces that assumption in its advice. This phenomenon is known as algorithmic bias: a form of prejudice that is not deliberately programmed, but stems from the data. In the history of artificial intelligence, bias in training data has proven to be a recurring stumbling block, from facial recognition to language models.

Why Is This Such a Serious Problem?

The risk of AI salary advice is greater than it might initially appear. A growing number of employees and job seekers — particularly younger generations — consult AI chatbots for practical career guidance. Think of questions like: "What is a realistic salary for a marketing manager with five years of experience?" or "How do I negotiate my salary?"

If ChatGPT systematically benchmarks women lower, they may enter negotiations with expectations that are already too low. The consequence: they ask for less, accept less, and the pay gap is not closed but actually reinforced. This is especially concerning because AI tools are often perceived as objective and neutral — a misconception that makes the bias all the more dangerous.

Moreover, a growing number of HR departments and recruiters are using AI tools for salary determination and job evaluation. If those tools carry the same bias, inequality seeps into the entire organization. You can read more about how AI is being deployed in the workplace on our page about AI applications.

What Is OpenAI Doing About It?

OpenAI is aware of the issue of bias in its models and has taken steps in successive versions of ChatGPT to reduce prejudice. The company uses techniques such as Reinforcement Learning from Human Feedback (RLHF) to correct undesired behavior. Yet the most recent research shows that the pay gap in salary recommendations is far from resolved.

Critics argue that OpenAI and other AI developers need to be more transparent about the limitations of their models, especially when those models are used for sensitive applications such as financial or career advice. They call for mandatory bias audits before AI tools are brought to market, similar to the requirements the European AI Act imposes on high-risk AI systems.

Tips for Users

Until AI salary recommendations become more reliable, there are several practical steps employees can take:

Use AI salary advice as a starting point, not a final verdict. Combine it with salary benchmarks from specialized platforms.
Frame your question without mentioning gender and consider comparing the answer with a version in which you include different demographic details.
Consult trade unions, HR professionals, or salary databases for a more objective picture.
Be aware that AI tools are not neutral — they reflect the data on which they were trained.

Conclusion: AI Reinforces Inequality If We Are Not Careful

The finding that ChatGPT systematically recommends lower salaries for women is a wake-up call for anyone who blindly trusts AI tools. Artificial intelligence offers enormous possibilities, but it also reproduces the blind spots and inequalities of the society that created it. As long as AI companies like OpenAI do not work in a structural and transparent way to eliminate gender bias, every woman who asks ChatGPT for salary advice remains potentially disadvantaged.

The technology of the future must not become an instrument that digitizes and entrenches existing inequality. Follow more AI news on stersoftware.com, or explore the opportunities and risks of artificial intelligence through our knowledge base.

BusinessWise