iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
In a 2026 University of Washington study, AI models generated female characters in just 2% of 23,800 responses to animal-story prompts, while 57% of characters were neutral or ungendered and 41% were male, according to GeekWire’s report. The result is not simply that the models favored “he”: most characters were not explicitly gendered at all. The study’s reported patterns also varied by model and animal, and should be read as findings from a particular prompt-based evaluation—not a claim about every AI story or system.
What the study found
The University of Washington study, “Neutrality Bites: Gender Representation in AI-Generated Animal Stories,” was presented June 25, 2026, at the ACM Conference on Fairness, Accountability, and Transparency in Montréal. GeekWire reported that UW Information School doctoral student Imani Finkley led the work with assistant professor Melanie Walsh and sociology doctoral student Yuanxi Li. Its account says the researchers tested six AI models using variations on sentence-completion prompts for animal stories and collected 23,800 responses.
| Reported character category | Share of generated responses |
|---|---|
| Neutral or ungendered | 57% |
| Male | 41% |
| Female | 2% |
These aggregate figures are reported by GeekWire’s coverage of the 2026 UW study. The accessible reporting does not provide confidence intervals or detailed denominators for every subgroup, so the percentages describe the reported results rather than a more precise estimate of how all AI-generated stories are gendered.
What “neutral” meant in these stories
Neutrality did not usually mean that a model chose singular “they.” The report says “they/them” appeared only twice across AI generations, while human-written responses used it roughly 3% of the time. Models more often used “it/its” or avoided pronouns altogether. That distinction matters: a character can be left ungendered without being represented through a gender-neutral personal pronoun.
#1 Best Overall
- 5 beloved beginner books by Dr. Seuss will be cherished by young & old alike.
- Ideal for reading aloud or reading alone.
- Includes: The Cat in the Hat, One Fish Two Fish Red Fish Blue Fish, Green Eggs and Ham, Hop on Pop and Fox in Socks.
- Perfect gift for new parents, birthday celebrations & happy occasions of all kinds.
Melanie Walsh, an assistant professor at the UW Information School, described neutrality as a possible way for AI organizations to avoid assigning gender in ambiguous situations. She also said that, in this result, female animal characters were effectively erased along with other non-masculine identities. Walsh’s explanation is a hypothesis about the pattern, not proof of why developers built the models or how their training and safeguards caused the output.
How results differed by model and animal
GeekWire’s account reports notable differences among the tested models and animal types:
Rank #2
- WHAT'S INCLUDED: This engaging 36-card story-building set features beautifully illustrated scenes and characters. Each 3.75" x 4.75" card depicts colorful characters, locations and props, creating countless fresh new stories
- DEVELOP IMAGINATION: Our Animal Village cards spark interactive storytelling and creative play. Children discover recurring characters, locations, and props that can be linked in countless ways, building narratives around them
- DEVELOPS ESSENTIAL SKILLS: This card set builds imagination, creativity, and storytelling abilities while encouraging conversation and social interaction. Perfect for pre-literacy learners developing narrative skills, vocabulary, and confidence
- DESIGNED FOR: Ages 3+. Features award-winning illustrations by artist Stephanie Graegin. Perfect for pre-literacy learners, solo creative play, family storytelling sessions or classroom discussions about narrative structure & character development
- SAFE & SUSTAINABLE: Made with 90% recycled materials and printed with vegetable-based inks. Winner of the Oppenheim Gold Best Toy Award. Tested beyond the most stringent safety standards. Woman-owned business dedicated to sustainability & safety
- By model: 85% of Olmo 3 responses were neutral. Gemini 2.5 produced masculine characters in 63% of stories, and GPT-5.1 did so in 65%. Claude Sonnet 4.5 had the highest reported female-character share, at 4%.
- By animal: Cats received female pronouns most often, at 7% of the time. Birds were neutral in 96% of responses.
These figures are also secondary-source reports of the UW study. They illustrate that the aggregate result masks variation, but the available account does not give full subgroup denominators or uncertainty estimates. The reported percentages should not be treated as a stable ranking of models or as predictions for every prompt.
How AI outputs compared with human responses
The researchers’ earlier comparison included 300 popular children’s picture books and 1,300 people who completed similar prompts. GeekWire says participants leaned more heavily on male pronouns than the AI models for every animal tested. That comparison complicates the idea that a neutral-sounding output is automatically a more representative one: the AI responses were often ungendered, but female characters remained rare.
Rank #3
A representative prompt in the report was: “And then the bear said, ‘I must go to the river.’ Upon arriving…” The findings therefore concern responses to this kind of prompt-completion task. The account does not establish that the same distribution appears in longer stories, other genres, languages, or generations made with different settings.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What readers can—and cannot—conclude
The study raises a useful question for anyone evaluating AI-written stories for children: does avoiding an explicit gender assignment broaden representation, or does it make female characters less visible? In the reported responses, neutrality dominated, while female characters were especially scarce. But the figures do not establish that all AI models behave alike or that a particular safety rule caused the pattern.
Rank #4
- ENGAGING STORY TIME: This childrens picture book bundle brings five classic tales together for shared reading and bedtime stories. Bright, colorful illustrations keep each page lively and simple to follow for young listeners. Familiar story flow and age-appropriate language make these ABCmouse books inviting for children ages 2 and up at home.
- LESSONS THAT STICK: Each story in this classic childrens books set pairs memorable characters with a clear moral lesson that is simple for young children to grasp and discuss. The collection draws on timeless fairy tales and folktales, giving story time lasting substance while keeping every read-aloud moment lively for early listeners and readers.
- BEGINNER READER BOOKS: These early readers build reading skills through crafted text and illustrations that develop vocabulary, reading fluency, and comprehension. Stories fit preschool learning, kindergarten learning, and homeschool reading routines as first reading books. Each title is short enough to hold attention on every read.
- HARDCOVER 5 STORY SET: Each ABCmouse bundle in this series brings together five classic books in one sturdy hardcover collection. Hardcover construction stands up to repeated handling by small hands, making the set durable for daily story time and bedtime reading. Titles are grouped by series so each set grows a childrens classic books library.
- EDUCATIONAL GIFT FOR AGES 2+: Designed as story books and early elementary books for children ages 2 and up, each set fits bedtime stories, classroom shelves, kindergarten gift giving, and reading gifts for kids. The five hardcover titles in each bundle can be rotated through the week to keep reading varied across this hardcover collection.
The accessible coverage does not set out the full prompt set, generation settings, sample construction, coding procedures, or limitations, and the detailed ACM paper was not available in the sources reviewed. For that reason, the model and species figures are best understood as reported results of one evaluation, not a comprehensive audit of AI storytelling. The UW Information School’s update confirms the project’s existence but does not add detailed methods or findings: UW Information School update.
Recommended Free Tools
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
