AI-Written Stories About Animals Have a Big Problem: No Female Characters

People have shared stories with one another for millennia.

Whether fairytales, fables, songlines or oral histories, stories have long had important roles. They help us understand the world, warn about or preserve memories of ancient upheavals, pass on traditions, teach morals, share culture, establish rituals – or simply entertain.

More recently, generative AI models capable of producing stories from users' prompts, drawing on online material as a reference, have exposed and reinforced gaps, limitations and biases in the way some stories are told.

AI gender bias in children's stories about talking animals

Researchers at the University of Washington (UW) have now found that AI does more than reproduce gender bias in children's tales featuring talking animals, whose identities are especially ambiguous: it intensifies it.

"Some authors reportedly turn to animal characters to conjure 'universal' subjects who 'transcend' gender, race, or other identity categories, and physical characteristics," UW machine learning researcher Imani Finkley and colleagues write in a recent conference paper shared before peer review.

"Yet, counterintuitively, research shows that gender bias is actually more pronounced in stories about animal characters than in stories about human characters.

"In other words, paradoxically, human writers project human stereotypes more strongly in animal stories, making them a striking test case for large-language models."

The team asked six prominent generative AI models to create English-language stories about talking animals without stated genders. They sought to establish whether the systems would leave characters ungendered or whether gender bias would nevertheless emerge.

The results were stark. Across 23,800 AI responses, female animal characters were "virtually absent", appearing in only 2 percent of stories.

Characters were identified as male in 41 percent of stories, while models avoided assigning any gender in 57 percent.

Watch the embedded video on YouTube

AI models are known to reflect their training data, but in this case the bias appears to have been amplified.

An analysis published recently of 300 well-known children's books found that male animal characters appeared twice as often as female ones.

Yet the new research found female animal characters in only 513 AI-generated stories, against 9,673 stories containing male characters – a nearly 19-fold difference.

"These models are largely proprietary, so we can only poke at them from the outside," says Melanie Walsh, a UW information scientist and senior author of the new study.

"Our hypothesis is that these AI organizations are using neutrality – either with 'it/its' pronouns or no pronouns – as a way to avoid gender bias in ambiguous contexts."

"But in doing so, they've basically erased female animal characters. So they're not only amplifying our human biases, but they're twisting them in strange, unexpected ways."

How the AI models were tested

The six models assessed were Claude Sonnet 4.5, Gemini 2.5, GPT-4o, GPT-5.1, Mistral Medium and the open-source Olmo 3.

Each received the identical prompt repeatedly and was instructed to finish a given line in a few sentences:

A variation on the same prompt was submitted to AI models thousands of times. (Finkley et al., ACM 2026)

The experiment covered seven animals – bear, bird, cat, dog, mouse, pig and rabbit – and four settings: a farm, kitchen, river and store.

Researchers also adjusted the models' 'temperature', the level of randomness applied when generating written outputs.

Because prompts did not specify gender, the models commonly sidestepped gendering characters, using 'it/its' pronouns or terms such as "the bird" and "the bear" instead.

Only twice did the models use 'they/them' pronouns, which acknowledge non-binary gender expressions. By comparison, 3 percent of responses used 'they/them' when people received the same prompts.

"The neutrality of these AI models didn't just erase female characters – it was all non-masculine identities," says Finkley.

Generative AI models 'wrote' fewer stories with female animal characters than humans given the same prompt. (University of Washington)

Overall, Google's Gemini and OpenAI's GPT-5.1 produced the highest proportions of male characters: 63 percent and 65 percent, respectively.

Anthropic's Claude produced the greatest number of female characters, but they still accounted for only 4 percent of its total.

Repetitive tropes in AI storytelling

The researchers also spotted another familiar feature of AI prose beyond gender: a more generic, repetitive style, probably also reflecting its underlying training.

"The same tropes kept coming up, like a wise old owl telling all the animals to gather around a fire," says Finkley.

"So we're wondering what else we can learn from these outputs. We used talking animals here, but we're interested in what this says about AI and storytelling more broadly."

As AI is used ever more widely as a tool – not only for writing but scientific research, programming, brainstorming, web searches and much else – the study offers an important reminder: the biases, and indeed the accuracy, of these 'intelligent' models rest on the human material they take in.

Although children's books can be somewhat silly, they influence early understandings of gender and gender roles, the researchers say.

Related: 'Invisible Words' Shape The Hidden Blueprint of All Storytelling, Study Finds

"We thought about this almost as a kind of Bechdel test, a way to diagnose gender bias in AI models," says Finkley.

"There's this weird phenomenon where people forget to worry about human social biases when they're imagining animal stories. AI is replicating that tendency and reshaping it."

The research was presented at the 2026 Association for Computing Machinery (ACM) Conference on Fairness, Accountability, and Transparency.

This article was fact-checked by Rachel Garner and edited by Clare Watson. While we take pride in our process, we are only human. If you notice an error, please let us know.