Fashion trend forecasting helps brands anticipate what styles will be popular in upcoming seasons. Traditionally, this process has depended on experts’ intuition, creativity, and experience. More recently, big data analytics has provided more profound insights into consumer behaviour. Still, these methods often require advanced technical skills, putting them beyond the reach of fashion students and small designers.
Advances in artificial intelligence (AI) are now helping to level the playing field. Large language models (LLMs) such as ChatGPT can analyse vast amounts of cultural and societal data, making data-driven insights more accessible to the public. However, since LLMs are prone to inaccuracies and “hallucinations,” researchers must carefully test their effectiveness in practical applications, such as fashion forecasting.
In a recent study published in the Clothing and Textiles Research Journal on 26 September 2025, Assistant Professor Yoon Kyung Lee and Master’s student Chaehi Ryu from Pusan National University explored how ChatGPT could be used to predict fashion trends. “Rather than simply asking what will be popular in the future, we designed a systematic method to prompt AI for specific and consistent answers,” explained Dr Lee. Their work compared ChatGPT’s predictions with reports from a professional trend forecasting agency.
To refine the process, the researchers created a new Top-Down Prompting (TDP) method inspired by the Lotus Blossom brainstorming technique. This method begins with a general prompt—such as “predict fashion trends”—and expands into detailed subtopics like silhouette, materials, key items, decorative elements, colour, and mood. Using this framework, they generated ChatGPT-3.5 and ChatGPT-4 predictions for men’s autumn/winter 2024 fashion and compared them with forecasts from the Official Fashion Trend Information Company (OFTIC), reviewed by fashion experts.
The findings showed that ChatGPT tended to reproduce familiar or mainstream fashion ideas rather than highly innovative designs. It matched only 9 of 39 trends in OFTIC’s report. Nonetheless, both models detected emerging cultural themes, such as gender fluidity and statement outerwear, indicating AI’s potential to sense broader social movements and inspire creativity.
Although ChatGPT is not yet reliable enough to replace expert analysis, it offers a valuable supplementary tool for education and small-scale designers. By combining AI insights with human expertise, the TDP method makes fashion forecasting more systematic, inclusive, and accessible—opening new opportunities for creative exploration and data-informed design.
More information: Yoon Kyung Lee et al, How the Field of Fashion can use ChatGPT to Predict Fashion Trends, Clothing and Textiles Research Journal. DOI: 10.1177/0887302X251371969
Journal information: Clothing and Textiles Research Journal Provided by Pusan National University