According to a report, 65 percent of Swedish consumers state they use AI tools regularly, with 26 percent using them daily. ChatGPT is the most used application (76 percent), followed by Google Gemini (43 percent).
Within e-commerce, 52 percent of consumers have used AI in the past year to conduct research or receive product recommendations. The corresponding figure for use in physical stores, such as via mobile during a store visit, is 20 percent. The most valued aspects in the physical environment are AI-based real-time price comparisons (39 percent) and product scanning for access to additional information (28 percent).
Skepticism Towards Autonomous Purchases via AI
Despite the high usage for information seeking, 44 percent of consumers are negative towards allowing independent AI agents to execute and complete purchases on their behalf. Only 20 percent are positive about the concept. The primary barrier is the need for control; 57 percent of negative respondents state they want to retain full control over their purchasing decisions.
However, many would feel more comfortable if critical parts such as payment and ID verification were handled separately and securely. The report indicates that fully autonomous purchasing systems for broad consumer use are likely to be delayed, and that future solutions will more likely require a human in the decision-making chain.
Frank Almenning, business developer at BankID, commented on the issue during a roundtable discussion in May 2026:
The technical aspects for payment through AI agents are largely already in place, but who bears the responsibility if something goes wrong?
New Standards and Data Containers
Behind the scenes, development is underway on the infrastructure to manage the future of agent-driven commerce. In early 2026, Google launched the Universal Commerce Protocol (UCP) standard to orchestrate purchasing processes and checkout systems for AI agents. In April 2026, actors such as Meta, Microsoft, Amazon, Stripe, and Salesforce joined the protocol’s technical advisory board.
Swedish startup companies are also building new services based on so-called data containers, where business and consumer data can be stored separately. The H&M-supported company eComID is developing digital shopping passes that contain body measurements and style preferences. This allows AI models to provide more accurate recommendations without storing the information themselves.
Shift Towards Semantic Search
For retailers, the development means that AI models are increasingly functioning as intermediaries between store and consumer. Review platform Trustpilot reports a sharp increase in traffic from AI models searching for structured and unstructured data to answer consumer product questions.
The development represents a shift from traditional keywords and SEO to semantic search, where AI models analyze deeper context and consumer actual experiences rather than individual words. According to Trustpilot, the models primarily focus on three criteria when evaluating information: relevance, ranking (trust signals), and timeliness. It is therefore becoming more difficult for companies to hide flaws in products, as AI models easily identify recurring negative patterns in aggregated data.