Disappointed, but not surprised. That was my first reaction when I read the new Nature study revealing that generative AI tools are amplifying one of society’s most persistent and overlooked biases: the disappearance of women and, increasingly, men, after the age of 40. This blog explores the question – Does AI have a gender bias?
What was surprising was the response my post received on LinkedIn, so it has to be worth further investigation, right?
So, the basis of my initial post was that the research found that large language models such as ChatGPT tend to portray women as younger and less experienced than men, even when the input data is identical. When asked to generate professional résumés, AI consistently produced versions for women that made them appear to have fewer years of experience, while men of the same hypothetical profile came out older, more senior, and more “hireable.”
For anyone who has watched how women over 40 fade from visibility in popular culture, workplace narratives, and even stock photography, the results feel hauntingly familiar.
AI Is a Mirror and a Magnifier
Artificial intelligence doesn’t invent bias from nowhere. It reflects what we, as a society, feed into it: the data, language, and images we collectively produce and consume.
When that collective culture already devalues aging, particularly in women, the algorithms trained on it inevitably reproduce that distortion. What’s more concerning is that AI doesn’t just reflect these patterns; it magnifies them, reinforcing stereotypes at scale and speed we’ve never seen before.
The study’s authors, Douglas Guilbeault and his colleagues at Stanford, Berkeley, and Oxford Universities, found that this age-gender bias is woven deeply into both text and imagery across the internet. Women are systematically portrayed as younger than men in depictions of higher-status or better-paid roles. And the same tendency shows up in the billions of words used to train language models.
In other words, invisibility isn’t an accident. It’s embedded in our data, our assumptions, and now, our machines.
The Disappearing Act of Midlife
This phenomenon, the quiet fading of women after 40, has been playing out for decades. You see it in film and television, where middle-aged women are often relegated to the background or written out altogether. You see it in hiring patterns, where “youthful energy” is valued over “seasoned experience.”
Now, AI has joined the chorus. When a system that underpins hiring platforms or résumé screening tools unconsciously rates older women as less qualified, it doesn’t just reflect bias, it mechanises it.
And while this affects women most acutely, it’s not limited to them. Men, too, are increasingly subject to the erasure that comes with an obsession with youth. What was once a social pattern is fast becoming a digital law of nature … unless we intervene.
Does AI have a gender bias? – Reversing the Bias
The problem is not unsolvable, but it demands awareness and action at multiple levels.
If AI learns from us, then we must become intentional teachers.
We can:
- Demand transparency in how AI systems are trained and what data they use.
- Broaden representation in datasets, ensuring that people of all ages, genders, and backgrounds are visible in the material that trains these models.
- Institute bias audits as a standard part of AI development and procurement.
- Challenge our own language and imagery, asking what assumptions are embedded in the words and visuals we choose to represent “competence,” “innovation,” or “potential.”
Because ultimately, AI is not an alien intelligence, it’s a cultural one. It learns from our collective imagination. If that imagination excludes or diminishes older women, then so will the technology built upon it.
Reclaiming Visibility
The promise of AI is not inevitability but possibility. These systems could help us see more, to uncover hidden stories, amplify underrepresented voices, and rebalance the narratives that shape our world.
But that will only happen if we take responsibility for what we’re teaching them.
The challenge is clear: ensure that in training our machines to understand the world, we don’t also teach them to unsee the people who built it.
AI should not narrow our vision. It should expand it. Hopefully we’ve helped you answer the question, does AI have a gender bias?
Author: Deborah Holmwood, Client Change & Transformation Partner.
Follow our LinkedIn company page to stay up to date with all our new blogs!
