The AI Stylist Arrives for the Holidays, and Fit Is Still the Hard Part

On October 5, thirty years after it launched gap.com, Gap Inc. put an AI stylist in front of its shoppers. Old Navy and Banana Republic customers can now describe what they want in everyday language and get products back, and two outside styling apps, Alta Daily and Daydream, now draw on the Gap Inc. assortment. Two days later, Levi's told analysts that shoppers who use its new AI shopping assistant add items to the bag at roughly three times the average rate.

The timing is deliberate. Adobe expects traffic from AI platforms to United States retail sites to rise 130% this holiday season, and YouGov finds that a third of likely Black Friday shoppers plan to use AI during the sales. Apparel brands want a voice in those conversations, so they are building stylists of their own rather than leaving the advice to a chatbot that has never seen their size chart.

The early numbers are encouraging, but they measure the moment a sweater goes into the bag, which in apparel is only the start of the story. The same research that shows shoppers flocking to AI also shows them regretting what it told them to buy, and clothing sits near the top of that list. This post looks at what Gap and Levi's launched, what the data says about trust, and why fit will decide whether the AI stylist pays its way.

1. Thirty Years After gap.com, Gap Hires an AI Stylist

Gap's announcement works on three layers, and each says something about where apparel discovery is heading. The first sits on its own sites. Old Navy and Banana Republic now let a shopper ask for, say, a warm outfit for a December wedding, instead of translating that idea into keywords and filters. Gap said the tools are meant to help people move from an idea to products that fit what they had in mind, which is a fair description of a good sales associate.

The second layer lives off Gap's sites entirely. Alta Daily, an AI closet and styling app, now builds outfits from Gap merchandise around a user's schedule, budget and local weather. It can mix those pieces with clothes the user already owns and show the look on the user's avatar. Daydream, the fashion search platform led by Julie Bornstein, takes requests in everyday language across all four Gap Inc. brands. In effect, Gap is handing its assortment to stylists it does not own, because it expects discovery to start somewhere other than its homepage.

The third layer is the plumbing, and it may matter most. In March, Gap became a launch partner for Bold Metrics' Agent Sizing Protocol, which lets a conversational agent ask about fit before it recommends a size. It also signed on to Google's Universal Commerce Protocol, which is meant to let shoppers buy Gap products inside Google's AI Mode and the Gemini app. The budget is growing to match, since capital spending will reach $650 million this fiscal year, up from $470 million, with part of the increase going to AI. Chief technology officer Sven Gerjets told Retail Dive that AI should "not add complexity or take choice away from the customer."

2. Levi's AI Shopping Assistant and the 3x Add-to-Bag Number

Levi's supplied the first number that looks like proof. On its third-quarter call on October 7, chief executive Michelle Gass described the new AI shopping assistant on levi.com, which offers styling advice and builds complete outfits. Shoppers who use it add to bag at roughly three times the average rate, she said, while calling the result early. The context explains the focus on the site: e-commerce revenue grew 10% in the quarter, while direct-to-consumer revenue grew only 2% and comparable sales were flat.

The figure deserves the caution Gass gave it, for two reasons. Shoppers who open a styling assistant are already more engaged than the average visitor, so they would likely add more to the bag with or without it. More importantly, add-to-bag sits several steps before the number a finance team cares about, which in apparel is revenue that stays sold after the return window closes. An assistant that turns one pair of jeans into a full outfit has raised the basket, but whether it has raised the margin depends on how many of those pieces fit.

The same call offered a reminder of what an assistant cannot do. Levi's back-to-school campaign leaned on loose-fitting denim just as shoppers moved toward low-rise styles, and the company spent the quarter chasing inventory in the fits that were selling. A stylist can only recommend what is in stock, and it cannot rescue an assortment that misread the size or silhouette curve, as the SKU diet post argued. Levi's plans a full redesign of levi.com for 2027 with what Gass called "immersive fit navigation", which suggests the company already sees fit as the next problem to solve.

3. Shoppers Use AI Stylists More Than They Trust Them

Adoption is not the constraint. A SmartCustomer survey of 1,181 United States consumers, fielded in August, found that 76% had used AI to shop in the past year. Half expect to use it more over the next twelve months. Trust lags well behind, since 89% do not fully trust AI's recommendations, and three in four check them at least some of the time against reviews and their own research.

The holiday data draws the same line between research and buying. YouGov found that 34% of likely Black Friday and Cyber Monday shoppers expect to use AI during the sales, up from 16% a year earlier on a differently worded question. Comfort drains away as the AI moves closer to the money. While 43% would let it find the best price, only 9% would let it add items to their basket and just 5% would let it complete a purchase after approval. Global Payments' 2026 Agentic Commerce Report, which surveyed more than 16,000 consumers in seven markets, found that only 21% would let an agent finish a purchase alone, and 42% worry it would buy the wrong item.

For apparel brands this is less discouraging than it sounds. Shoppers want help deciding and want to keep the final say, which is exactly the role a brand's own stylist can play: an adviser that knows the size chart and the fabric, with the shopper still pressing the button.

4. Where AI Stylists Go Wrong: Fit and the 45% Problem

Regret is the number apparel brands should watch most closely. One in three shoppers in the SmartCustomer survey had made a purchase on an AI recommendation that they later regretted. Electronics accounted for half of those regretted purchases and clothing and shoes for 45%, well ahead of beauty at 33%. The most trusting group, millennials aged 30 to 44, was also the most likely to have been burned.

Apparel's place on that list is not surprising, because a clothing recommendation has to be right on things a chatbot cannot see. Style is the easy part. Fit depends on the shopper's body, the brand's grading and how a particular fabric drapes, and Coresight Research found that size and fit are the leading reason for apparel returns, cited by 53% of the brands and retailers it surveyed. An assistant that picks a size from a generic chart is guessing. In a category where about a quarter of online sales already come back, as the DoorDash returns post set out, a confident guess is an expensive one.

This is why the plumbing in Gap's announcement may matter more than the conversation. A sizing protocol that asks about fit first moves the hardest question to the front of the purchase, where it is cheapest to answer. Adobe's survey of 5,000 shoppers points to the prize, with 69% of those who have shopped with AI saying they are less likely to return what they bought. That answer is self-reported and sits uneasily beside SmartCustomer's regret figures, yet the two describe the same fork in the road. AI advice that gets fit right should lower returns, while advice that gets it wrong will raise them at scale.

5. Add-to-Bag Is Not the Metric That Matters

If AI stylists are going to be judged, they should be judged on what stays sold. Engagement, add-to-bag rate and conversion all stop at checkout, which flatters any tool that makes the basket bigger. In apparel the honest measure is kept revenue, meaning what an AI-assisted order is worth once the return window closes, net of the cost of handling what came back.

Getting there means splitting the numbers in a way most brands do not yet do. Return rates should be compared between orders where the shopper used the stylist and orders where they did not, by category and by size, since an assistant that does well on outerwear and badly on denim will hide its weakness in a blended average. Outfit recommendations need the same scrutiny, because a complete look is a bigger basket and also more chances for one item not to fit. A shopper buying one style in two sizes shows up at the order level, not in a SKU report.

The other half of the experience is what happens after the order. A Trustpilot study reported by WWD found that AI appears in only 0.3% of footwear reviews, yet those reviews are the angriest in the category, and almost all describe a chatbot standing between the shopper and a refund. Trustpilot's reading was that shoppers want AI to help them discover and decide, and a human when something goes wrong. A brand that puts a charming stylist in front of the purchase and an unhelpful bot behind it will spend the goodwill it earned.

6. Five Moves for Shopify Apparel Brands Before Black Friday

Most direct-to-consumer brands will not build a Levi's-scale assistant before the holidays, and they do not need to. The more useful work is making sure that whichever AI the shopper consults, whether the brand's own, a styling app or a general chatbot, has what it needs to give good advice. With AI-referred traffic expected to more than double this season, that preparation pays even for a brand with no assistant at all.

First, write product data for a stylist to read. Fabric composition, garment measurements, fit notes, the model's size and how a style runs against the brand's other fits are what an AI needs to pick the right size. Many product pages still leave them out or bury them in images.

Second, put fit into the reviews. Shoppers check AI advice against reviews more than any other source, so review prompts that ask about size, height and fit create the evidence that careful shoppers and careful AI both look for.

Third, measure AI-assisted orders on their own. Tag the orders where a stylist or an AI referral was involved, then track their return rate, exchange rate and kept revenue against everything else.

Fourth, keep a human behind the bot. Use AI where shoppers welcome it, in discovery and in the decision, and make sure a return or a problem order reaches a person in a few clicks.

Fifth, treat the order as the last styling decision. An AI that recommends a full outfit has made a bet on fit across several items at once. Checkout is the last point at which a brand sees that whole basket before anything ships.