65% of Shoppers Will Use AI This Christmas. 8% of Retailers Feel Ready.

The Gap, in One Pair of Numbers

Narvar published its 2026 Holiday Shopping Survey on 20 August, run with eTail Insights across 1,348 shoppers and 100 retail decision makers. Two findings sit right next to each other:

• 65% of shoppers plan to use AI for at least one part of their holiday shopping.
• 8% of retailers feel very confident in their AI capabilities.

A gap that wide is not a difference of opinion about the future. One side has already changed its behaviour and the other has not finished discussing it.

The detail underneath is sharper than the headline. Asked what most influences conversion, 52% of retailers named shipping costs and free shipping. That is not a foolish answer. It was the right answer for a decade, and free delivery still moves a checkout. But it is an answer about the last step of a journey whose first step has quietly moved somewhere else. If the shopper never reaches your product page, your delivery proposition never gets read.

This Is Measured Behaviour, Not a Forecast

Six days later, on 26 August, BCG published research drawn from more than 13,000 consumers across 12 markets, over 100 in-depth interviews and an analysis of more than 1,000 brands. The numbers describe something that has already happened:

• 31% of consumers now use AI somewhere in the purchase journey, roughly triple the rate of 18 months earlier.
• 19% are "AI loyalists" who rely on it regularly.
• Of those who use it, 70% buy the product the AI recommended.

That last figure is the one to sit with. Being named by the agent is not a soft branding win somewhere near the top of a funnel. For seven in ten of those shoppers, being named is the sale.

"Many companies are still competing for a consumer who has already moved on," said Aparna Bharadwaj, Managing Director and Senior Partner at BCG.

The Finding That Should Interest Smaller Merchants Most

Buried in the same BCG research is a result that runs against the usual assumption about who benefits when a new channel opens.

More than half of AI-assisted purchase journeys introduced the consumer to a brand they would not otherwise have considered.

Think about what that means mechanically. Traditional search is weighted towards incumbency: domain authority, backlinks accumulated over years, brand-name queries typed by people who already know you exist. A shopper who searches "Nike running shoes" has already decided most of the question.

An agent asked "what should I get my dad for Christmas, he cycles to work in the rain and hates being cold, under £50" has no incumbent to default to. It has a specification. It selects the product whose data most convincingly matches that specification, and in more than half of cases the answer is a brand the shopper had never heard of.

That is the most favourable discovery mechanic a smaller merchant has been handed in twenty years. It is also conditional, and the condition is unglamorous: the agent can only establish fit from data it can actually read. A product page that a human would find persuasive and an agent cannot parse loses to a plainer product with better structured data every time.

Why the Season Makes This Sharper

Gift buying is unusually well suited to agents, for three reasons that all peak in December.

The query is a specification, not a product. People buying for themselves often know the item. People buying for someone else describe a person and a budget. That is intent-shaped input, which is precisely what keyword search handles badly and an agent handles well.

Availability decides the recommendation. In December, stock moves hourly. An agent that recommends something out of stock has embarrassed itself in front of its user, and these systems learn to avoid sources that do that. Live price and stock stop being hygiene and start being the difference between being cited and being dropped.

The deadline is real. Most of the year, "we will get to our product data next quarter" costs you an unknown amount of nothing in particular. Between now and December it costs you a specific, dated, once-a-year peak.

What Actually Has to Be True

The work that closes this gap is not a replatform, which is the main reason it is still achievable in September:

• Structured product data an agent can retrieve directly, rather than a page it has to interpret.
• Intent mapping, so each product carries the questions, problems and use cases it answers, in the words a shopper would use.
• Live price and stock, so a recommendation is still true at the moment it is made.
• Discoverability in the formats the major systems actually consume, rather than hoping a crawler infers the right thing.

A well-structured catalogue of 300 products is more useful to an agent than a badly structured one of 300,000. Scale is not the constraint here, and that is exactly why the 8% figure is an opportunity rather than only a warning.

Where Agenticfeed Fits

Agenticfeed does this specific conversion. It takes an existing catalogue, from Shopify or a Google Merchant Center feed, and produces the structured, intent-rich version agents need: every product mapped to the questions, problems and use cases it answers, with grounded answers and live price and stock, published where the major AI systems can reach it through a Google Merchant Center feed, an Agentic Commerce Protocol feed and an MCP connector.

For how an agent decides between two similar products, see How AI Shopping Agents Choose Products. For why optimising for agents is a different discipline from SEO, see GEO Is Not SEO.

Frequently Asked Questions

How many shoppers will actually use AI this holiday season?

Narvar's 2026 Holiday Shopping Survey, run with eTail Insights across 1,348 shoppers and 100 retailers and published on 20 August 2026, found that 65% of shoppers plan to use AI for at least one part of their holiday shopping. Separately, BCG's August 2026 research across more than 13,000 consumers in 12 markets found 31% already use AI somewhere in the purchase journey, roughly triple the rate of 18 months earlier.

Does AI shopping only benefit big, well-known brands?

The opposite, on BCG's evidence. More than half of AI-assisted purchase journeys introduced the consumer to a brand they would not otherwise have considered. An agent selects on fit with the stated need rather than on brand familiarity, so a smaller merchant with well-structured product data can be named ahead of a household name that an agent cannot parse.

Is it too late to prepare for this holiday season?

No, because the work is data preparation rather than a platform migration. What an agent needs is structured product data with live price and stock, and each product mapped to the questions, problems and use cases it answers. A focused catalogue of a few hundred products can be made agent-readable in days rather than quarters.

The Takeaway

The two studies were published by different firms, six days apart, measuring different things. They agree anyway. Shoppers have moved, most retailers know they have not, and the ones still treating this as next year's project are preparing carefully for the previous season.

The encouraging half is that the door is open wider for unknown brands than it has been in a long time. More than half the time, an AI-assisted purchase ends with a brand the shopper had never considered. Whether yours can be one of them comes down to something entirely within your control: whether a machine, asked a specific question by a real buyer in December, can read your catalogue well enough to name you.