Shopify's 2026 Global Holiday Retail Report, published on 6 October, found that 65%(Shopify Newsroom) of shoppers say they will use artificial intelligence for at least one shopping task this season, and that 71% of regular AI users are more likely to choose an independent brand than they were last year. The report arrives alongside OpenAI's launch of a virtual try-on feature inside ChatGPT and Adobe's projection that AI-referred traffic to US retail sites will rise 130% year over year. The practical question for merchants is no longer whether agents will send them traffic, but whether an agent can understand what they sell well enough to recommend it.
Key Conclusions
The first conclusion is that product data quality has become a demand-generation function. An agent cannot recommend a product it cannot parse, and it cannot justify a recommendation it cannot verify. Shopify reports that 81% of regular AI users say they are choosing more products this year because those products feel personal, intentional or considered, which means the deciding factor is often the richness of the product record rather than the size of the advertising budget behind it.
The second conclusion is that AI referral rewards brands with a specific, describable customer problem. Cottonique, a hypoallergenic apparel brand, saw AI-referred sales grow 276%(Shopify Newsroom) year over year with total sales up 52%, and Stanley 1913 recorded five times as many AI-referred orders in the United States. Neither result came from a keyword campaign. Both came from describing clearly what the product solves, for whom, and why the claim can be trusted.
What the Report Measures
The report measures three things that together explain the shift. The first is adoption: 65% of shoppers intend to use AI for at least one shopping task, and Adobe separately projects a 130% increase in AI-referred traffic to US retail sites. The second is trust, which remains partial: only one third of shoppers say they would trust an agent to complete a purchase on their behalf. The third is behaviour, where 35% of shoppers choose the best possible product for their needs even when it is not the cheapest option, which is precisely the kind of decision an agent can support well.
Discovery without a keyword
Agentic search changes the input. A shopper can describe a need in their own words, such as a fabric that irritates sensitive skin, and an agent that understands the underlying problem can surface a brand that solves it without requiring the shopper to know the industry term. That removes the historical advantage of brands that invested in keyword coverage and returns it to brands that invested in explaining a genuine product benefit clearly enough for a machine to repeat.
The trust gap that remains
The gap between research and transaction is still wide. Only one third of shoppers will let an agent buy on their behalf, so most AI-influenced purchases are still completed by a person who was persuaded earlier in the journey. That has a direct implication for content strategy: the material an agent reads has to carry the argument, because the agent will summarise it faithfully, and a thin product page produces a thin recommendation.
Best Practices
Merchants preparing for agent-mediated discovery should focus on three actions. First, structure product data into explicit attributes so an agent can compare items on the same dimensions rather than guessing from marketing copy. Second, state the customer problem in plain language near the top of the record, because that is what an agent matches against a request. Third, place the same claims in authoritative third-party sources, since agents weigh corroboration heavily when deciding whether a brand can be trusted.
Write for comparison, not only for browsing
Agents compare. A product page that describes an item in isolation gives an agent nothing to work with, while a record that states material, dimensions, care requirements, warranty terms and intended use gives it comparable fields. Adding a virtual try-on experience, as OpenAI has now done inside ChatGPT, raises expectations further: shoppers can visualise a garment, which means the accompanying data has to be accurate enough to survive that level of scrutiny.
Let the assortment be findable in more than one place
A retailer that only exists in its own storefront limits its own discoverability. The same week, a department store expanded its toy assortment and added new app features for the holiday period, and a mass retailer continued to push its own assistant deeper into search and recommendations. Brands that appear across an owned site, a marketplace and an AI channel give an agent more opportunities to verify and repeat their claims, which is what converts a single mention into a durable recommendation.
Common Mistakes
The first mistake is assuming that search engine optimisation translates directly into agent visibility, when the underlying requirement is structured, verifiable product facts. The second is writing product copy for human persuasion while leaving attribute fields empty. The third is ignoring third-party corroboration and relying solely on self-published claims. The fourth is treating AI traffic as a channel to be bought rather than a record to be maintained. The fifth is failing to measure AI-referred revenue at all, which makes it impossible to know whether any of the work is producing a return.
The Measurement Gap Behind AI Referrals
Most merchants cannot currently say how much revenue arrived through an AI assistant, because the referral is often indistinguishable from a direct visit or an organic search. That measurement gap matters more than it first appears, because it prevents any feedback loop from forming. A retailer that cannot attribute the sale cannot learn which product records performed well, cannot justify further investment in structured data, and cannot defend the budget when other channels compete for it. Closing the gap, even approximately, is a prerequisite for treating agent visibility as a managed channel rather than an accident.
Summary
The first agentic holiday season is testing a capability that most merchants have never had to build deliberately: legibility. Shopify's finding that 65% of shoppers will use AI for at least one shopping task, combined with Adobe's forecast of a 130% rise in AI-referred traffic, describes a discovery layer that sits between the shopper and the brand. What that layer repeats depends entirely on what the brand has written down in a form a machine can read.
The work is unglamorous and measurable. Fill the attribute fields, state the customer problem plainly, publish the same facts in independent sources, and instrument AI referrals so the effect can be seen. Brands that do this will be found by a machine and chosen by a person, which is exactly how the report frames the outcome of the season.
Data Sources
- Shopify Newsroom: Welcome to the first holiday season of the agentic era (6 Oct 2026)
- Retail Dive: Virtual try on launches on ChatGPT (9 Oct 2026)
- Retail Dive: 2026 holiday e-commerce sales growth forecast (28 Sep 2026)
- Retail Dive: AI is reshaping Walmart and Sparky is leading the charge (8 Oct 2026)
- Retail Dive: Nordstrom expands FAO Schwarz offerings and app features (10 Oct 2026)
- NRF: 2026 retail sales forecast
FAQ
What does it mean for a brand to be legible to an AI agent?
A: It means the product record states attributes, intended use and verifiable claims in a structured form, so an agent can compare items and repeat the brand's argument without inventing detail.
Why does AI referral help smaller independent brands?
A: Because agentic search matches a described need rather than a keyword, a niche brand that genuinely solves a specific problem can be surfaced without outbidding larger competitors for search terms.
Does a virtual try-on feature change what product data is needed?
A: Yes. Once shoppers can visualise a garment, fit, material and care information must be accurate enough to survive closer inspection, otherwise the visualisation amplifies a return rather than preventing one.
How should merchants measure AI-referred traffic?
A: Tag assistant referrals where possible, compare conversion and return rates against other channels, and track which product records appear most often in assistant answers.
What is the single highest-return action to start with?
A: Completing the attribute fields for the top-selling assortment, since agents compare on structured dimensions and empty fields make a product effectively invisible in comparison.
References
- Welcome to the first holiday season of the agentic era
- Virtual try on launches on ChatGPT
- Consumers aren't just buying gifts this holiday season. They're stocking up on essentials.
- AI is 'reshaping' Walmart and Sparky is leading the charge
- Nordstrom expands FAO Schwarz offerings, debuts new app features for holidays
- NRF Forecasts
