What happened

Narvar released its 2026 Holiday Shopping Report on 24 August. In its survey, 65% of consumers said they planned to use AI for at least one part of holiday shopping. Among retailer respondents, 8% described themselves as very confident in using AI to improve the shopping experience.

The research covered 1,348 US consumers planning to shop online and 100 retail decision-makers. It measured stated intentions and confidence, not completed purchases or a global adoption rate.

The useful question for a retailer is how well its products and service promises survive an AI-assisted shopping journey.

Why it matters

A shopper asking for a suitable gift may express needs rather than a product name: a budget, an age range, a delivery deadline or a particular interest. A retailer should be able to answer those questions clearly in its own information before expecting another system to interpret them well.

Consider a hypothetical gift with several variants. If dimensions are missing, compatibility is vague and delivery terms contradict one another, an assistant may struggle to make a dependable comparison. The same gaps can frustrate a human customer.

I would begin with a small group of commercially important products. Check the facts a buyer needs to make a decision: what the item does, who it suits, what is included, what it costs and when it can arrive. Assign owners to keep those facts consistent across relevant channels.

This is useful work regardless of which shopping interface becomes popular.

The bigger shift

Discovery and fulfilment need to be considered together. An attractive recommendation is of limited value if the stock position is wrong or the delivery promise cannot be met.

A practical readiness exercise can follow five ordinary shopping questions. Ask about a product comparison, a compatibility issue, a budget constraint, a delivery deadline and a return condition. Test the answers against the retailer’s approved information.

Record missing facts and unsupported claims. Repeat the exercise across relevant interfaces without treating any one result as a guaranteed ranking. Different questions, locations and systems may produce different answers.

Then follow the journey through purchase and support. Make sure the customer can verify an important promise and find help when an answer is wrong. Track actual enquiries, completed orders and service problems alongside visibility.

The existing guide to AI search optimisation explores the discovery side. Operational readiness also requires accurate information and a service team able to honour it.

My take

I would use the survey as a reason to inspect the shopping experience, rather than as a forecast to paste into a revenue plan.

The first useful deliverable is a list of information defects, ranked by their effect on a purchase decision. Fix the most consequential ones, then repeat the same test. That provides evidence of progress without requiring an expensive new platform.

Keep the commercial measurement separate. If customers begin using an AI-assisted route, examine whether it brings appropriate demand and whether those customers receive what they expected.

Retailers do not need to predict precisely how every shopper will choose a gift. They do need to make their offer understandable, their availability credible and their promises consistent. Those are strong foundations for both human and AI-assisted discovery.