Buying with AI in Italy: what the Shopify report says about Christmas 2026

The Shopify report on Christmas 2026 describes the stated intentions of Italian consumers with respect to AI in shopping. The most cited data, 72%, it concerns the probability of using AI for at least one buying activity: searching, comparing or evaluating products is not the same as delegating the payment yourself. Factsheet Italy.

Shopify published the Global Holiday Retail Report on 6 October 2026. For an Italian e-commerce, the useful reading is preparing comprehensible and consistent product information in the different channels. The report does not show that a similar percentage of sales will go from autonomous agents and does not announce a new checkout available everywhere. Shopify Announcement.

Before changing investments, it is better to understand what was asked, to whom and when. Then you can translate the signal into concrete work on catalog, content, assistance and measurement. A survey of intentions helps to formulate hypotheses; the results of one's shop require observed data.

What does the Italian AI data really mean?

In factsheet Italia, the 72% declares the use of AI likely for at least one shopping activity, the 47% is said to be comfortable with product recommendations and the 36% trusts a delegate purchase. Formulations measure different levels of availability and are not interchangeable. Official Italian data.

A consumer can use an assistant to compare features and then complete the purchase on the site, after checking price and delivery. Another may accept a recommendation but not authorize a payment. Presenting everything as an autonomous purchase erases these differences and leads to too strong a commercial reading.

The report then offers a signal on the interest towards assistance in the path. For the store, the immediate question is whether the information needed for that route is available, up-to-date, and understandable. The Guide to the AI for e-commerce helps link this work to catalog, customer care and internal processes, without reducing it to a new sales button.

How was the Italian investigation carried out?

The Italian consumer sample includes 2,000 adults, interviewed online at July 2026. The Business Survey covers 1,000 decision makers, interviewed ad August 2026, in companies with less than 1,000 employees. The research is attributed to Sapio Research. They are samples and different periods, not to be merged. Methodology in factsheet Italy.

The company population described does not automatically coincide with the definition of microenterprise or with the typical customer of a small Italian store. The e-commerce manager should therefore avoid applying each conclusion to his or her context without further verification. Industry, assortment, and audience can influence how customers use AI tools.

The website of the Global Holiday Retail Report gathers the general picture and local materials. To cite an Italian behavior you have to use the relevant Italian document. No independent verification of sampling is carried out here, nor are sales volumes inferred from questionnaire replies.

What numbers should not be mixed?

The table does not describe a store conversion funnel. Percentages cannot be multiplied to derive expected sales or interpreted as consecutive shares of the same process. They are answers to different questions, to be read with the methodology and with the exact meaning of the formulations.

Even the global data, 65%, does not replace the Italian one. A title on the national market must use the correct local reference. Shopify Global Report. Accuracy matters because a sentence about intentions can quickly become, in the transition between presentations, a commercial prediction never demonstrated by the source.

What does an assistant need to find in the catalog?

A product should have recognizable name, specific description, consistent attributes, distinct variants, and updated information. Sizes, measurements, materials, and compatibility must be expressed unambiguously. If a feature changes between page, feed, and management software, an assistant can receive conflicting signals and the customer can make a decision on the wrong basis.

There is no need to rewrite every description in promotional form. It is often more useful to make the limits explicit: what the product includes, what an accessory requires, when it is not suitable. The guide to AI prompt for e-commerce can help prepare content, but features must come from approved sources.

Recurring questions of care are a good track. If customers often ask for compatibility or return conditions, that information must be easy to find and consistent. The work on the catalogue does not guarantee visibility in every AI engine; however, it makes the material that customers, staff and tools can consult more solid.

Hypothetical scenario: a spare parts shop prepares the catalog

Let's imagine an Italian spare parts e-commerce that wants to reduce the ambiguity of the cards before the commercial season. The team compares codes, compatible models and variants between management software and site. Where a technical confirmation is missing, it does not ask the model to complete the card: it puts it in review with the product manager.

Test applications include requests made in common language and with names not identical to those in the catalogue. The team checks whether it is possible to find the correct product and understand the limits of compatibility. It is not only necessary to generate an answer: it is necessary to identify the source and make the information verifiable.

This is a hypothetical scenario. It shows no additional sales nor a position in AI results. The test observes data quality and customer path. If an information is inconsistent, the fix must reach all relevant channels, preventing an updated description on the site from coexisting with a still-wrong feed.

Hypothetical scenario: an accessory brand prepares Christmas

Consider an accessory brand that offers gift products. The team wants to make materials, dimensions, packaging and delivery conditions understandable. Instead of assuming that customers will delegate payment to an agent, it improves answers to questions that precede the decision and verifies the transition to the product page.

Creativity can use variants produced with Higgsfield Ad Multiplier, while page design can be explored with Figma AI. They are activities distinct from the report: neither a generated creativity nor a new screen demonstrate an increase in conversions.

This case is also hypothetical. The criterion is to maintain consistency between the ad, the product page and actual availability. A delivery promise must not be invented to make the message more persuasive. The team prepares an update procedure for changes in stock or conditions, keeping the content connected to operational work.

How to measure paths without inventing attribution?

The store must distinguish visits, requests, purchases and quality of care. A customer can consult an AI and get to the site through a path that analytics does not fully identify. The absence of a recognizable source does not prove the absence of influence; likewise, a visit from an AI service does not prove that the sale was caused by that service.

Recognisable entrances can be observed and compared with behavior on the site, keeping explicit limits. Questions to staff and customer requests can add context, without automatically becoming a statistical measure. The goal is to formulate verifiable assumptions, not to attribute each sale to the newest channel.

The novelties of ChatGPT Ads on Formats and Measurement They relate to another level of the path. Also there you have to distinguish registered event, qualified lead and commercial result. The Shopify report does not provide an expected acquisition cost for individual e-commerce nor a guaranteed campaign return.

What verification to do before rewriting all tabs?

Open the tab for a catalog check

Choose a family of products and compare site, feed and management software. Check attributes, variants, compatibility, availability and conditions. Prepare realistic questions and check if answers are supported by sources. Correct the original data first, then update the channels that use it. Record the ambiguities yet to be clarified.

A reasoned sample helps to determine whether the problem is systematic. If the codes or variants are inconsistent in multiple tabs, a modification of the data model might be needed instead of a manual rewrite. AI can assist the comparison, but the decision on the correct data must remain tied to a reliable source.

The staff must also be able to maintain the result. A once-cleaned catalog may return inconsistent if the insertion process does not change. It is advisable to define mandatory fields, responsibilities and controls before publication. Useful work is what remains upgradable when new products arrive, not just what improves a photograph of the site.

Where is it wrong to read a report on AI purchases?

The first mistake is calling autonomous shopping assisted shopping activities. The second is to turn declared intentions into observed sales. The third is mixing Italian and global data. The fourth is to describe the business sample as representing only micro-enterprises, ignoring the perimeter indicated by the research. Method and local data.

Another error is to deduce from the report the availability of each channel in Italy. Announcement of search, technical integration and activation on your account are different evidences. Before designing a new checkout or sales via agent you need to verify product, market, condition and real operation.

Finally, a measure of interest in AI does not justify generic mass-produced content. If product information sheets do not answer product questions, adding text can make the catalogue noisier. The work should reduce ambiguity and improve decisions, avoiding unverifiable promises and identical descriptions.

How to assign work between marketing and operations?

Marketing can identify questions and messages, but attributes and availability require the contribution of those who manage the product. The technical team must ensure that the information reaches the relevant channels. Customer care can signal what customers don't understand. A manager should coordinate these contributions and decide which source prevails in the event of a conflict.

The Guide to the AI for businesses proposes to start from the process. Here the process is the production and maintenance of information that supports a purchasing decision. The report can direct attention, while the project must solve observable problems in the store and in the team's work.

How to handle questions that the catalog does not solve?

A complete catalogue does not eliminate every request for clarification. A customer may describe an unanticipated use or ask about compatibility the supplier has not confirmed. The process should indicate how to collect these questions and who can answer. The assistant must not turn missing information into a positive recommendation.

Recurring requests can become a work of catalog improvement. The team identifies the theme, checks the response with an authorized source, and decides where to enter it. A new information should update the relevant tab, assistance and channels, while maintaining consistency. If it concerns only a particular case, it must be presented with that limit.

Reviews also require distinction. A customer opinion may help to understand the user experience, but it does not replace an approved technical specification. A synthetic content should not invent reviews or testimonials to fill a page. The goal is to make facts verifiable and recognizable opinions, without mixing them into the same commercial promise.

For complex products, it may be helpful to maintain an assisted path to staff. The client finds the information available and knows how to ask for further verification. This path must also be visible from mobile and consistent with the company's real-world ability to respond. A generic contact invitation is not enough if the question requires precise data.

The intention report offers a reason to observe these steps, but the priority comes from the problems of the store. If customer care is to continually clarify the same attribute, that is a concrete point to be improved. The decision does not require predicting how many future sales will be autonomous: it requires a correct source and a process that keeps it available.

What concrete improvement to start from?

Choose an assortment where questions and features are important to the decision. Test sources, correct inconsistencies, and test the path with realistic demands. Measure what you can observe and state the attribution limits. The preparation also applies when the customer completes the purchase with a traditional route.

If your catalog changes often and information diverges between site and internal systems, work Giallo e-commerce can start with the alignment of data and cards. It is a concrete way to prepare for AI-assisted research without turning an intent survey into a sales promise.