Social Rocket
AI may help with product search – but many people still want to control the binding purchase themselves. In a German survey for the KPMG Consumer Snapshot 03/2026, 79 percent say they want to be able to intervene in a shopping agent's process at any time and decide for themselves. 73 percent can imagine using an agent to compare products, prices or offers.
For German online retail and social commerce ahead of the Christmas season, this sets a concrete priority: those using AI for recommendations, gift finders and advertising links should weight transparent product information and a controllable checkout higher than fully automated purchases.
The decisive line runs between selection and payment
The study, published by KPMG on 8 October 2026 and analysed by the EHI Retail Institute, illustrates the difference using a smartphone purchase: 72 percent would use AI support when searching for information; when it comes to the payment process, the basic willingness drops to 45 percent. Both figures describe stated attitudes in a survey scenario, not actually completed AI purchases.
The usage figure also needs to be read correctly: 46 percent said they had used AI for purchase decisions frequently or very frequently in the past twelve months. That is neither a market share of autonomous agents nor a conversion rate.
The German perspective complements the US Feedvisor study on AI shopping, in which 62 percent reject autonomous orders. The studies use different surveys and questions; their percentages must therefore not be combined into a shared global rate.
Trust is not built through maximum automation
Alongside the right to intervene, according to KPMG/EHI 74 percent want to decide for themselves which personal data the AI uses. 72 percent want to be able to set rules and budget limits; 67 percent want transparency about whether provider interests influence a recommendation. On perceived objectivity, 64 percent prefer general generative AI over offers from retailers or manufacturers; only 7 percent favour the latter. This is a statement of preference, not an independent quality comparison of the systems.
For brands, this leads to four verifiable requirements: current prices and availability must be easy to find, recommendations should state comprehensible criteria, affiliate and sponsorship relationships must remain recognisable, and an order should offer a clear approval step before payment.
The practical foundation is often less spectacular than the AI agent itself: clean product data, reliable variants, understandable return and shipping terms, and precise details on delivery areas. Relatedly, Google is expanding AI visibility and commerce tools in Merchant Center. However, these product features are not automatically available in Germany.
Global AI optimism is not German purchase willingness
The Shopify study on the holiday season 2026 reports that 65 percent worldwide want to use AI for at least one shopping task. This does not contradict German reluctance around payment: support in searching and handing over purchase authority are two different tasks.
In addition, Shopify's figure is a statement of intent for the coming season. Anticipated AI use is not a measured order and says nothing about the incrementality of sales. The US beta "Ask Target" meanwhile shows how a retailer positions AI first for inspiration and product selection, rather than handing every order over to an autonomous agent.
What marketing teams should measure before Black Friday
A meaningful test separates AI-assisted research, cart building and confirmed purchase. Look at more than clicks from AI and social channels: also measure actual orders, cancellations, returns, margins and the number of manual corrections needed for price or availability. High engagement with the gift finder may simply mean a new step in the existing buying process – not additional revenue.
Social media teams in particular should ensure consistency in creator recommendations and AI answers: product name, current promotion, link destination and the promoted variant must match. If you cannot guarantee this, AI may end up causing more abandoned purchases rather than fewer.
Conclusion: The German findings do not argue against AI in retail. They argue for an assisted, verifiable buying process in which people can use recommendations without giving up control, data sovereignty and budget limits.
Sources: KPMG/EHI Consumer Snapshot 03/2026 from 8 October; Gartner, US surveys from 27 May; Shopify Holiday Report, 6 October.