No title card, no logo in the corner, no end frame, no typeface of your choosing. A dress turns inside a rounded card, one of up to eight in a row, under three lines of text. That is what a garment film looks like once it answers a shopper’s question inside ChatGPT instead of opening a campaign. The film stops being advertising and becomes an interface component — cut to be legible in seconds, inside a frame the brand does not own.

None of that is an aesthetic preference. OpenAI’s UI guidelines for apps inside ChatGPT tell developers to “keep to 3–8 items per carousel for readability,” hold metadata to “three lines max,” permit “a single, optional CTA” per card, and forbid custom fonts “even in full screen modes.” Imagery “must follow enforced aspect ratios to avoid distortion.” The rule a creative director should read twice is blunter: “Do not include your logo as part of the response. ChatGPT will always append your logo and app name before the widget is rendered.”

Read that as a deletion list. The title sequence, the logo bug, the bespoke type, the end card, the house colour grade: every device the fashion film has used for a century to announce who made the clothes is now either banned or supplied by the platform in a form the brand cannot touch. ASOS Stylist, which put shoppable video inside ChatGPT in May, arrives under a logo OpenAI stamps on, set in OpenAI’s font, cropped to OpenAI’s ratio. What survives the trip is the garment and the body moving in it.

The clothes now carry the entire burden of identification, unassisted.

Fashion has filmed under these terms before. Pathé’s cinemagazine Eve’s Film Review carried fashion as discrete segments inside newsreel programming, close-ups favouring material detail, garments shown in sequence rather than story, to audiences who had bought a ticket for something else. When colour was the selling point and black-and-white could not carry it, Pathé stencil-tinted the print: “For Afternoon,” 1922, billed as “A Pathécolor Study.”

The design problem is the same one, and it forces legibility early. A campaign film can spend four seconds on an empty room before anyone walks in, because the viewer chose to watch it. Answering ASOS’s own example prompt, “show me pastel floral A-line dresses for spring,” a clip competes with up to seven neighbours in a scroll strip, declares the silhouette in its first beat, and has to survive a small card with the sound off. That pushes the grammar back toward the stencil: high contrast, one garment, movement starting on frame one.

The commercial case for bothering comes mostly from the company that built the thing. Bambuser, which supplied the video layer behind ASOS Stylist, published a survey of 1,000 UK and US consumers on 9 June finding 78% of Americans more likely to buy after watching a product video, and half of shoppers leaving brand sites mid-session to hunt for “real life” video on TikTok or YouTube. A vendor’s research into the value of the vendor’s product deserves the obvious discount. The figure that is harder to wave off is smaller: 19% of US consumers now name ChatGPT or AI search as a way they discover products, against 21% for brand websites.

The trade filed this as a distribution story, another surface to push assets onto, which is the part already solved. Look at what the guidelines actually legislate: font, colour, card count, aspect ratio, the number of buttons a shopper may press. They say nothing about how long a turn should last, whether a clip should loop, or where the cut falls. The visual grammar of the answer engine is being written now, by whoever ships first — currently a fast-fashion retailer in Camden and a video company from Stockholm.