AI assistants send few shoppers to fashion sites, but they now buy more

In the two months before Christmas 2025, traffic from generative AI tools such as ChatGPT, Gemini and Perplexity to US retail websites rose 693% year on year, according to Adobe Analytics, which tracks more than one trillion visits to US retail sites. On Thanksgiving, visitors arriving from an AI assistant converted 54% better than everyone else.

The short answer to the question fashion executives keep asking is this: AI assistants are now sending shoppers who buy, but not many of them. Bain, working with Similarweb, found that AI accounts for up to a quarter of referral traffic for some retailers, yet still less than 1% of total retailer traffic. Zalando, Europe's largest online fashion platform, told investors in March 2026 that chatbot traffic was a "very low single percentage points" share of its organic traffic.

It matters now because the quality of the traffic has flipped faster than its volume. In early 2025 AI visitors were the weakest converters on retail sites; by mid 2026 they were the strongest, while growth rates were falling from four digits to double digits. For fashion, which Adobe ranked among the lowest converting categories for AI traffic in early 2025, the window to prepare product data, content and checkout for this channel is open, but not indefinitely.

How fast is AI traffic to retail sites growing?

Very fast, but the curve is bending. Adobe first flagged the trend in March 2025, when it reported that generative AI traffic to US retail sites in February 2025 was up 1,200% on July 2024, doubling roughly every two months since September 2024. Over the 2024 holiday season the increase had been 1,300% year on year. The 2025 holiday season brought a further 693%, with November up 769% and December up 673%.

Since then, the year-on-year comparisons have become harder. Adobe measured growth of 393% in the first quarter of 2026, 138% in May 2026 and 62% in July 2026. Measured against October 2024, when Adobe began tracking the series, AI referral traffic in July 2026 was up 1,219%. May 2026 set a new monthly high, above every month of 2025, even without a seasonal lift.

Similarweb, which measures referrals across the open web rather than inside a retailer's analytics, sees the same pattern at global scale. In the twelve months from June 2025 to May 2026, AI platforms sent an average of 770.7 million referral visits a month to websites worldwide, up 117.4% on the previous year. Fashion referrals grew 278.2%, faster than news or travel, rising from 4.4 million visits in May 2025 to 6.7 million in October 2025. Beauty grew 312.5%. Marketplaces were the biggest retail winner, averaging 46.8 million AI referrals a month.

Similarweb counts only visits that carry referrer data, so its figures are a floor. Visits from apps or copied links can land in the "direct" bucket. That caveat cuts both ways: the true volume is somewhat higher, but nobody can say by how much.

Behind the traffic sits a shift in habits. In Adobe's survey of more than 5,000 US consumers, published in March 2025, 39% said they had already used generative AI for online shopping and 53% planned to do so that year. The most common tasks were research (55%), product recommendations (47%), finding deals (43%) and gift ideas (35%), all early-funnel activities in which fashion brands traditionally relied on search, social media and magazines to be discovered.

Bain's Consumer Lab put the share of US consumers using generative AI for product research and comparison at 30% to 45% in November 2025. Ahead of the 2025 holidays, 17% of online shoppers told Bain they would start their holiday shopping on an AI platform such as ChatGPT or Perplexity, against 30% who planned to start with Google search, itself now infused with AI answers. The generational split is wide: 52% of millennials and 25% of Gen Z said they planned to begin holiday shopping with an AI assistant. For fashion, whose growth depends heavily on younger customers, those intentions matter more than today's small traffic shares.

Do shoppers from ChatGPT convert better than other visitors?

On Adobe's data, yes, and the reversal is the most striking number in this story. In July 2024, AI visitors to US retail sites were 43% less likely to convert than other traffic. By early 2025 the gap had narrowed to 9%. Over the 2025 holiday season AI referrals converted 31% better than other sources. In March 2026 the premium was 42%, in May 2026 it was 54%, and in July 2026 it reached 60%, the eleventh consecutive month in which AI traffic outperformed.

Engagement tells a similar story. In July 2026, Adobe said AI-referred shoppers spent 59% more time on site, bounced 33% less, added items to the basket 28% more often and generated 53% more revenue per visit than other shoppers. The revenue picture used to be the opposite: during the 2024 holiday season, non-AI visits were worth 51% more per visit than AI visits.

Salesforce, which aggregates data from more than 1.5 billion shoppers, reached a comparable conclusion for the 2025 holidays. It said visitors from AI-powered search converted nine times more often than visitors from social media, and that the share of traffic arriving from third-party AI search channels doubled year on year. Salesforce also estimated that AI and agents influenced USD 262 billion of global online sales over the season, around 20% of the total, although its release does not define "influenced".

Why do some studies say AI traffic converts worse?

Because they measure different things. A working paper by Maximilian Kaiser (University of Hamburg) and Christian Schulze (Frankfurt School of Finance and Management), reported by Modern Retail in October 2025, analysed twelve months of first-party data from 973 e-commerce sites with combined annual revenue of USD 20 billion. It found ChatGPT drove more than 90% of large language model traffic to those sites but less than 0.2% of total traffic, roughly 200 times less than Google organic search. On a last-click basis, affiliate links were 86% more likely to convert than ChatGPT referrals and organic search outperformed ChatGPT by about 13%.

The authors wrote that their "results contradict widespread expectations of LLM superiority", while noting that ChatGPT conversion rates rose month by month during the study period. The paper has not yet been peer reviewed. Adobe's sample is different: large US retailers using its analytics, more than 200 of the Top 2000 online retailers with over USD 836 billion of 2025 e-commerce sales between them. Large, well-known merchants appear to benefit most when an assistant recommends a brand by name.

Which fashion brands are AI assistants recommending?

Mostly the big ones. Similarweb's AI visibility leaderboard for fashion in the US, based on ChatGPT answers in September 2026, puts Nordstrom first with a visibility score of 25%, followed by Macy's (15%), Old Navy (14%), Uniqlo and Target (12% each) and H&M (10%). The most cited domains in those answers include department store sites and fashion publishers such as Who What Wear and Vogue, a reminder that assistants lean heavily on editorial roundups.

Traffic data point in the same direction. Similarweb found that ChatGPT accounted for more than 80% of AI referrals to the world's top 1,000 domains in mid 2025, so visibility in a single assistant still decides most of the traffic. Marketplaces, which combine broad assortments with structured product data, captured the largest retail share of AI referrals, averaging 46.8 million visits a month between June 2025 and May 2026, up 237.3% on the previous year. Monthly marketplace referrals rose from 33.2 million in June 2025 to 89.9 million in May 2026.

What did Zalando tell investors about chatbot traffic?

Zalando offers the clearest disclosure from a large fashion platform. On its fourth quarter 2025 call in March 2026, co-chief executive Robert Gentz said the company typically ranks "1, 2, or 3" in chatbot answers to fashion and lifestyle queries, according to a MarketBeat summary of the call. Referral traffic from chatbots was "growing a lot", he said, but remained "a very low single percentage points" share of organic traffic and "not yet very material".

Zalando's response has been to treat assistants as an "incremental channel" rather than a threat to its own app. Gentz said Zalando was "doubling down" on its in-house team for in-app AI experiences, was working with an AI lab, and was "one of only two companies" Google works with on its Universal Commerce Protocol for agentic shopping. He also described a "technical and logistics consolidation layer" through which Zalando could help brands sell via agentic channels.

What happened when Walmart let shoppers buy inside ChatGPT?

The most instructive case so far comes from outside fashion. In November 2025 Walmart made about 200,000 products available through OpenAI's Instant Checkout, which let users pay without leaving ChatGPT. In March 2026 Daniel Danker, Walmart's executive vice president of product and design, described the in-chat experience as "unsatisfying". According to MarTech, purchases completed inside ChatGPT converted at roughly one third of the rate of sessions in which users clicked through to Walmart's own site.

Walmart's next step is to embed its own assistant, Sparky, inside ChatGPT, with shoppers logging into their Walmart accounts and checking out in Walmart's system; a similar integration with Google Gemini is expected. OpenAI, MarTech reported, has begun phasing out Instant Checkout in favour of merchant-controlled checkout. The lesson for fashion brands is that the assistant may own discovery, but the retailer's own checkout, account and returns logic still close the sale better.

Better than most categories, but not by enough. Adobe's AI Content Visibility Checker found that in July 2026, 76% of apparel retailers' homepage content was machine-readable, the highest of any category, ahead of electronics (70%) and cosmetics (68%). Across all page types in May 2026, however, apparel was only 51% readable, behind cosmetics (63%), electronics (56%) and sporting goods (51%). Adobe said "consistent and structured product content is helping drive AI visibility" in apparel, while warning that many US retailers "continue to have AI visibility gaps".

That matters because fashion was a weak performer early on. In its March 2025 report, Adobe said conversion of AI traffic was highest in electronics and jewellery and lowest in apparel, home goods and grocery. Clothing is harder to recommend from text alone: fit, fabric and stock by size change constantly, and an assistant that cannot read size availability or returns terms will hesitate to send a shopper.

What are the risks and limits of the AI traffic data?

• Tiny base. Growth of several hundred per cent on a base below 1% of traffic is still a small channel. Bain's figure of less than 1% and the academic finding of less than 0.2% should temper any reallocation of budgets.
• Vendor data. Adobe, Salesforce and Similarweb sell analytics and AI services. Their methods are disclosed only in part, and Salesforce does not define what "influenced" sales means.
• Attribution gaps. Referrer-based counting misses visits from apps and copied links, while last-click models ignore assistants that inform a purchase made later via search or directly.
• Selection effects. AI referrals may convert well because they come from desktop-heavy, higher-income users: Adobe found 86% of AI traffic was desktop in early 2025, and that high-income US states generated 52% of AI traffic.
• Thin evidence. Juozas Kaziukėnas, an independent e-commerce analyst, told Modern Retail: "There's so much noise about AI agents impacting commerce and so little data." Assistants that now send free referrals may also later charge for placement.

There is also a consumer trust ceiling. Bain found that about half of US consumers are not comfortable letting AI handle purchases end to end, and that consumers trust retailers' own on-site agents three times more than third-party agents. Christian Schulze, co-author of the Hamburg and Frankfurt study, put it bluntly to Modern Retail: "If people don't really trust ChatGPT yet, then having Instant Checkout is not a meaningful feature."

None of this means the channel can be ignored. The pattern resembles the early years of search and of social commerce: a small share of visits, a disproportionately valuable audience and a short period in which early movers can establish positions that later become expensive to buy. The difference is that recommendation inside an assistant is far more concentrated than a page of search results, so being third or fourth in an answer may be worth much less than being first.

What should fashion companies do now?

1. Measure AI referrals separately. Create a dedicated channel group for ChatGPT, Gemini, Perplexity and Copilot referrals, and track conversion, basket size and returns against search and social.
2. Make product data machine-readable. Publish structured data for price, size availability, materials, care and returns terms on product and category pages, the content Adobe says drives AI visibility.
3. Earn editorial citations. Assistants draw on publishers and roundups, so press, reviews and inclusion in buying guides now feed discovery directly.
4. Keep checkout in your own system. The Walmart test suggests sessions that click through to the retailer's site convert better than purchases completed inside a chatbot.
5. Test your brand in assistants. Run a fixed set of category prompts monthly and record whether, where and how your brand is named, alongside competitors.
6. Size the opportunity honestly. Plan for a channel under 1% of traffic today, growing more slowly than in 2025, and avoid shifting budget away from search and owned channels on growth rates alone.