Last weekend, I was testing out the AI prompt to draw me a technology Barbie doll. I decided to follow it up with a male version.
I tested a few different AI tools to see what results I would get.
Adobe Firefly
I kept the default settings in Adobe Firefly. It created an image using Gemini 2.5 (with Nano Banana) in landscape mode. I used the prompt “Draw me a male Ken Barbie doll that works in Information Technology.” I was reasonably happy with the result, but he looks like he works in both IT support and fixes optic fibre.
I add the phrase “Remove the box and toolbelt.” The result was better. I even noticed the whiteboard had “server migration plan” written on it.
I noticed I wasn’t using Firefly Image 5 (preview). It’s a Premium feature, and the licence I have lets me try it now with up to 5 complimentary generations. When I tried swapping to Firefly Image, I got an error message: “Prompt declined. One or more words may not meet User Guidelines. Try changing your prompt.” I had to remove the words Barbie and Ken. The result was cute but not quite what I was going for.
Google Gemini
I was intrigued to know if Google’s Gemini would produce the same image with the same prompt as Adobe Firefly. I selected Thinking mode, and it used Nano Banana Pro. Ken has slightly changed his attire. He’s wearing brown glasses and swapped his jeans for khaki pants. He’s wearing the same light blue long-sleeve shirt and a grey vest. Perhaps it’s a year later, and he’s gotten a well-deserved promotion. His hair is slicked up. He has his own desk next to a server rack and a desk sign that reads “Ken – IT Analyst”.
For the record, I have never seen any guy in Brisbane wear a vest, let alone a grey one!
Grok Imagine
Grok rewrote my basic prompt to something more descriptive: “A male Ken Barbie doll dressed in professional Information Technology attire, realistic plastic doll style with perfect features, wearing a button-up shirt, tie, ID badge, and holding a laptop, standing pose, bright and clean studio lighting.” The results are squeaky clean: a crisp white shirt, slick hair, and a plain tie.
I don’t know anyone in the Brisbane tech industry who dresses like this. The only exception was a well-paid consultant who came from Sydney. I told him, “Nobody in Brisbane wears a tie, unless they have an interview”. He came back with a chuckle, “I like to keep them guessing!”
Conclusion
It is interesting to see how AI reflects and amplifies society’s existing biases. If you ask your AI program to generate a scientist, you’ll typically get a male in a lab coat. If you ask for a receptionist or a nurse, you’ll typically get a female.
We must challenge some of these typical stereotypes and ask who is building these tools, and what data is training them?
