In the fast-paced world of fashion, anticipating the next big trend is as crucial as designing the clothes themselves. Historically, fashion trend forecasting has been the domain of experts who rely heavily on intuition, a deep understanding of cultural shifts, and seasoned creativity. With the advent of big data analytics, there has been a paradigm shift toward integrating quantitative insights into forecasting, enabling brands to tap into consumer behavior patterns on a massive scale. Yet, this transition comes with significant hurdles, particularly for those operating on limited technical capacity, such as students and independent designers. Enter artificial intelligence (AI), a technology poised to democratize access to predictive analytics in fashion like never before.
Large Language Models (LLMs), like ChatGPT, have emerged as powerful tools capable of distilling enormous volumes of textual information spanning societal trends, cultural narratives, and consumer sentiments. By learning from diverse datasets, LLMs hold promise for application in industries outside traditional data-centric realms, including fashion forecasting. However, despite their vast potential, these AI models are not without their flaws. Known issues such as hallucinations—AI generating plausible but incorrect information—and difficulties with factual accuracy remain challenges. Thus, researchers have emphasized the importance of structured methodologies to harness AI effectively for reliable forecasting.
In a pioneering study from the Department of Clothing and Textiles at Pusan National University, South Korea, Assistant Professor Yoon Kyung Lee and Master’s student Chaehi Ryu explored innovative ways to utilize ChatGPT to predict fashion trends systematically. Unlike direct querying techniques that elicit broad and often inconsistent predictions, the researchers crafted a top-down prompting strategy. This method engages the AI with a hierarchical series of increasingly specific prompts rather than a single overarching question, thereby seeking precision and stability in the AI’s output. Their research, published in the Clothing and Textiles Research Journal in September 2025, sheds light on the intersection of AI and fashion forecasting, offering valuable insights for both academia and industry.
The study began by assessing ChatGPT’s baseline capabilities in predicting men’s fall/winter fashion trends for 2023. Initial general prompts helped calibrate the AI models’ behavior, revealing tendencies to generate generalized, established fashion concepts rather than cutting-edge predictions. Building on these preliminary findings, the researchers developed the Top-Down Prompting (TDP) technique, inspired by the Lotus Blossom brainstorming method, to divide the complex challenge of trend forecasting into manageable sub-components. This method prompts the AI to focus on discrete aspects of fashion such as silhouette, materials, key items, garment details, decorative elements, colors, moods, and patterns—each treated as a sub-problem to refine forecasting accuracy.
Applying the TDP technique, the team instructed ChatGPT versions 3.5 and 4 Classic to predict men’s fashion trends for the fall/winter 2024 season. The endeavor was methodically validated against the Official Fashion Trend Information Company’s (OFTIC) authoritative report on the same season, with further review from established fashion experts. This comprehensive approach ensured that AI-generated forecasts were not assessed in isolation but rather contextualized within the established expert knowledge and industry standards.
The comparative analysis demonstrated that while ChatGPT’s trend predictions were largely reflective of well-known and recurring fashion motifs, the models did identify emerging themes that were corroborated by expert observations. The AI accurately matched 9 out of the 39 key trends highlighted in OFTIC’s forecast. Remarkably, it also flagged pioneering concepts related to gender fluidity and statement outerwear—ideas that were gaining traction but not yet dominant in mainstream reports. Such insights suggest that AI has the potential not only to replicate existing trends but also to sense subtle cultural shifts that might inspire future design innovation.
Despite these promising signs, the study underscores that current LLMs should not be considered standalone forecasting tools. The relatively low accuracy rates in capturing comprehensive trend profiles highlight intrinsic limitations in AI’s interpretive faculties and data constraints. However, by complementing traditional expert-led analyses with AI-driven perspectives, the fashion industry can harness a hybrid approach that leverages the strengths of both human creativity and computational power. For smaller brands and educational institutions, especially those with constrained resources, this balanced strategy could democratize access to sophisticated trend analysis.
Key to the practical adoption of AI in fashion forecasting is the educational value of such tools. Recognizing this, the Pusan National University researchers have proposed a TDP-based hybrid conceptual framework that integrates AI-generated data with expert insights. This framework offers a systematic pedagogical model to teach fashion students how to effectively collaborate with AI technologies, enhancing their ability to generate nuanced and informed trend forecasts. The implications extend beyond education, potentially transforming how fashion design professionals approach the concept-to-market lifecycle.
The implications of this research reverberate through the broader landscape of fashion technology. By providing a clear, replicable prompting strategy, the study addresses one of the significant barriers to AI utility in creative industries: consistency and specificity in output. The Top-Down Prompting mechanism acts as a scaffold for engaging AI models with complex prediction tasks, ensuring that responses are structurally relevant and grounded in segmented thematic areas. This methodology could be adapted and extended beyond fashion into other creative and consumer-centric fields such as interior design, product innovation, and marketing strategy development.
Moreover, this research exemplifies a practical step toward mitigating AI’s known challenges—hallucinations and factual inaccuracies—by embedding verification and expert cross-referencing into the forecasting process. It highlights the necessity of human oversight in the AI forecasting pipeline, reinforcing the principle that AI should be viewed primarily as an augmentative tool rather than an autonomous oracle. This philosophy aligns with emerging trends in responsible AI deployment, emphasizing transparency, interpretability, and human-in-the-loop frameworks.
In conclusion, the study from Pusan National University not only validates the potential of AI-driven methods like ChatGPT for fashion trend forecasting but also provides practitioners with a tangible, structured approach to harnessing these tools effectively. It opens promising avenues for fashion students and small brands to leverage cutting-edge technology to anticipate market shifts, innovate design concepts, and compete in a highly dynamic industry. As the boundary between fashion creativity and computational intelligence continues to blur, such pioneering work lays the groundwork for an exciting future where AI empowers a more inclusive and innovative fashion ecosystem.
Subject of Research:
Fashion trend forecasting using AI and large language models, specifically ChatGPT.
Article Title:
How the Field of Fashion can use ChatGPT to Predict Fashion Trends
News Publication Date:
September 26, 2025
Web References:
https://doi.org/10.1177/0887302X251371969
References:
Lee, Y. K., & Ryu, C. (2025). How the Field of Fashion can use ChatGPT to Predict Fashion Trends. Clothing and Textiles Research Journal. https://doi.org/10.1177/0887302X251371969
Image Credits:
Yoon Kyung Lee from Pusan National University
Keywords:
Artificial intelligence, Data analysis, Machine learning, Marketing, Technology, Marketing research, Big data, Data sets
