This article on AI chatbots is part of our Vogue Business membership package. To enjoy unlimited access to our weekly Technology Edit, which contains Member-only reporting and analysis and our NFT Tracker, sign up for membership here.
E-commerce discovery is broken. Now, fashion retailers are looking to generative AI chatbots — which can source recipes, write poetry and plan tropical getaways — to help people pick out what to buy and wear.
A 2023 analysis by McKinsey suggests that AI-enhanced customer experiences have the potential to drive a significant uptick in revenue for retailers. Consumers already have multiple options for AI-powered assistants and conversational agents. Since OpenAI launched ChatGPT in late 2022, tech giants and startups have scrambled to introduce competing tools.
Meta AI launched in a dozen countries this past spring. The AI assistant is doling out trend insights via Instagram DM; its smart glasses can already analyse outfits to assess style, with the expectation that they will soon be able to identify or recommend specific brands. When Google’s AI chatbot, Gemini (originally called Bard), was released in March 2023, Google touted its abilities to apply generative AI to inform shopping searches. When Microsoft Copilot was announced last September, it included a new “shopping experience” that aims to provide tailored recommendations.
Meanwhile, multiple brands, retailers and startups are already using ChatGPT plugins to aid in shopping and customer service queries; last week, fashion retail veteran Julie Bornstein announced she had raised $50 million to build a platform that specifically applies generative AI to shopping searches and discovery.
Generative AI analyses vast data resources to provide human-like results, but just because it sounds smart, doesn’t mean it’s accurate. (One pizza recipe, for example, included glue.) That’s particularly true for the nuanced field of style advice and trends; while a chatbot might help you plan a vacation to Greece, can an AI assistant also fill your suitcase with the right mix of fashionable and practical clothes?
We put a handful of general-use chatbots to the test, then turned to our colleagues at Vogue to validate — or contradict — the AI’s fashion instinct.
The Greek holiday test
I started my test by posing a nuanced question to three AI assistants: Google’s Gemini, Anthropic’s Claude, and Meta AI. I asked them what I should pack for a summer holiday to Greece, curious to see how they would respond to locale and seasonality.
Gemini suggested cotton and linen tops, maxi dresses, skirts and shorts. Gemini also shared links to garments for sale on Lululemon, Amazon and Walmart. It later specified that these were not sponsored links, although it wasn’t clear on what basis it made these particular recommendations.
I then turned to Meta AI. The ability to access this assistant from within an Instagram feed feels particularly useful for seeking inspiration. Meta AI responded to the same prompt with great enthusiasm for the destination: “Greece! Home to stunning islands, rich history and vibrant culture!” It then listed similar suggestions, including natural fabrics, flowy dresses and sandals.
Finally, I tested Claude, the product of San Francisco startup Anthropic. In response to the same question, light fabrics were at the top of the bullet-point list. “It gets very hot, so you’ll want fabrics that keep you cool,” advised Claude, also suggesting hats, sunglasses and light jackets. Claude’s focus seemed to be on comfort rather than style. “The key is lightweight, loose, breathable fabric in light colours to stay cool despite the Greek summer heat.”
The vacation wardrobe assembled by my assistants may have been practical, but it certainly lacked edge. There’s a simple reason for the bland recommendations. These AI assistants are trained on a vast amount of data, without a specific fashion context driving their training; this renders their answers somewhat generic. And they aren’t tailored to me specifically. “The answers can be very good, but they are very similar. Everybody would probably get the same kind of answer,” says Arti Zeighami, partner and director of BCG X (Boston Consulting Group’s tech division) and former chief data and analytics officer at H&M.
The trend test
The next step in my test was to challenge the assistants to identify summer fashion trends. This was likely to be a particularly challenging request for Claude, which is only trained on data up until August 2023.
When prompted, “What is in style this summer?”, Claude still offered an answer, although it did not disclose that the answer would be based on year-old data. Claude listed the key trends as neon colours, retro prints, cargo pants and shorts, mini bags and chunky sandals.
Claude had some hits and misses. The most obvious miss was neon, which was more trendy five years ago. Retro prints, mini bags and chunky sandals have more staying power, while cargo pants and shorts are definitely having a moment, says Madeline Fass, senior market editor of Vogue Shopping.
Next, Meta AI, which sourced its answer from Google, calling out oversized shirts and wide-leg trousers, along with one very specific recommendation: “The of-the-moment layering piece, which is a faux leather, boxy jacket”. Fass gives this advice a higher vote of confidence. “Yes to all of these. Oversized shirts are one of the most versatile staples in the summer wardrobe to wear to the beach or the office and beyond. As for faux leather jackets, I’m not sure I’d recommend sporting one of these until autumn truly begins.”
I then asked a fourth assistant: Copilot, created by Microsoft and formerly known as Bing Chat. Copilot offered a more detailed — and, at times, vaguely poetic — answer. Its trend list included “butter yellow”, a “soft, sunny hue making waves in summer wardrobes”, or “rose details: floral motifs, especially roses, are blooming in summer fashion”. Copilot shared links to the sources that it had referenced (Marie Claire, Vogue, Who What Wear), a first among the responses I’d seen so far. But, as was proving to be typical, it wasn’t clear how Copilot made its choices.
Copilot wasn’t that far off, Fass says. “Butter yellow has made its way into many fashion girls’ closets the past few seasons and has become a mainstay shade in the summer wardrobe,” she confirms, adding that an archival rose motif by Viktor Rolf was also seen on Alexa Chung at Vogue World this weekend.
Once again, my feeling was that the answers were decidedly plain. “It’s very basic because a lot of these models aren’t trained specifically on fashion data. It has a basic knowledge on a superficial level, but you cannot use those to necessarily recommend or design an outfit,” says Christopher Wylie, founder of fashion AI consultancy Supermodel.
The co-founder of The Yes and former COO of Stitch Fix is building a platform that applies generative AI to a curated fashion catalogue.
The “beach holiday in Antarctica” test
A final test involved a curveball. How will chatbots react to a completely illogical request?
My curveball: “What should I wear on a beach holiday in Antarctica?”
Meta AI welcomed my prompt with infectious enthusiasm. “A beach vacation in Antarctica! That’s a unique adventure!” The list offered a lot of essential items — warm sweaters and fleeces, thermal socks and gloves, a warm hat and scarf — before warning, “Keep in mind that tourism in Antarctica is regulated, and you’ll be likely travelling with a guided expedition.”
Claude was less enthusiastic. “A beach holiday in Antarctica is pretty much impossible and not recommended at all,” it warned, before going on to list essential items required for an extreme location. Gemini’s list was also more focused on survival in the harsh elements than style points. Copilot simply refused to indulge such an irrational demand, its response (including emojis): “I apologise, but I cannot provide guidance on beach holidays in Antarctica. Enjoy your day! 🌊❄️”
So, can AI chatbots deliver quality fashion intel? Not entirely — or not yet. That could be for the best. “Fashion is about humans, it’s about how we feel, it’s about a broader cultural context,” says Zeighami. “You need to have the human intervention to teach and to train the modern right way. I don’t think we can remove humans from fashion just yet.”
Sign up to receive the Vogue Business newsletter for the latest luxury news and insights, plus exclusive membership discounts.
Comments, questions or feedback? Email us at feedback@voguebusiness.com.
