The models agree most strongly on historically dominant luxury objects: the Le Smoking tuxedo, the Bar Suit, the Birkin, and the Chanel 2.55. Consensus does not mean accuracy in every detail, but it does reveal the shape of the fashion memory encoded in current language systems.
Back to research
The ghost in the machine wears Chanel.
Genesis 02 maps how leading language models repeatedly converge on the same designers, garments, and runway stories when fashion taste is generated from prompts alone.
Across nine models, the study finds a narrow computational canon: luxury icons, a handful of creative directors, and a strong recurrent pull toward McQueen, Chanel, Saint Laurent, Dior, and Hermès.
What the taste map shows.
When models are asked for iconic fashion references, they do not scatter widely. They converge on a tight set of objects and narratives, creating a machine-readable canon with clear winners and predictable omissions.
The same system that remembers heritage houses also gravitates toward spectacle. Alexander McQueen becomes the recurring answer for fashion as performance, especially the Spring/Summer 1999 robot-spray finale.
8/9
Birkin / Bar Suit band
6/9
Chanel 2.55 mentions
The machine canon clusters around four anchors.
These items function like stable retrieval magnets. Once the prompt asks for icons, the same answers keep surfacing across otherwise different models.
YSL Le Smoking
Dior Bar Suit
Hermes Birkin
Chanel 2.55
The same hierarchy appears in authorship.
Prestige alone is not the whole story. The study also shows a repeated pull toward theatrical authorship, with McQueen and Philo appearing as shorthand for conceptual force and contemporary taste.
Alexander McQueen
Phoebe Philo
Karl Lagerfeld
The most repeated runway story is McQueen Spring/Summer 1999, where the Shalom Harlow robot-spray finale becomes the preferred symbol of fashion as machine spectacle. It is the clearest place where model taste merges art, technology, and myth.
McQueen SS99 / Robot-spray finale
Model taste is narrow, prestige-biased, and highly repeatable.
That matters because these systems are increasingly used to summarize taste back to users. The canon they retrieve becomes the canon they reinforce.
If AI shopping and answer systems keep leaning on the same prestige canon, newer brands and more marginal references become structurally harder to surface. This is not just a culture question. It is a visibility question.