# AI Shopping Agents Struggle With Real-World Purchase Accuracy Visa study reveals AI agents fail nearly half of consumer shopping tasks. - **Source:** [Fortune](https://fortune.com) - **Published:** 2026-11-05 - **Topic:** [technology](https://brifing.si/c/technology) - **Original article:** - **On Brifing:** https://brifing.si/s/9e07ceae-34e0-42d5-a401-59d62bcfa299 ## AI Shopping Performance Gap A recent study by Visa tested 22 different AI agents on their ability to purchase everyday items like TVs and laptops. The results revealed a significant performance gap, with an overall success rate of only 57%. This indicates that current AI technology still struggles with the complexities of autonomous consumer shopping tasks. ## Top Performing AI Models The study highlighted that cutting-edge models achieved the highest success rates. Google's Gemini Aragon 4 and Meta's Making Wonder both reached an 82% success rate. Meanwhile, an OpenAI agent ranked sixth with 64%, and a Claude agent secured third place with 76%, demonstrating that model sophistication directly correlates with shopping task completion. ## Cost Versus Success Correlation Visa analyzed the relationship between the cost of running an inquiry and the agent's success. The five most expensive models achieved an average success rate of 67%, while the five cheapest models only managed 50%. This suggests that higher computational resources currently play a vital role in improving AI decision-making accuracy. ## The Challenge of Ambiguity Agents often failed because they could not handle incomplete consumer information. In the experiment, agents had to identify products based on partial requirements, such as gaming console compatibility or specific display features. Success required the agents to actively elicit missing details from the shopper, a skill that many models currently lack. ## Future of Agentic Commerce Rubail Birwadker, Visa’s Global Head of Growth, stated that while the technology is not yet ready for full trust, it is advancing rapidly. The industry is now focusing on product discovery as a foundational element. Companies are working to integrate these agents into mainstream commerce through improved payment and security infrastructures. ## Payment and Security Hurdles Beyond product selection, companies like Visa and Mastercard are addressing the complexities of secure payments for AI agents. Facilitating so-called agentic commerce requires new tools that ensure transactions are safe and reliable. Retailers and technology firms are currently collaborating to define the standards necessary for this emerging digital shopping ecosystem. ## Unresolved Legal Liability Issues The rise of autonomous shopping agents introduces significant legal questions regarding liability. It remains unclear who is responsible if an agent makes rogue purchases or commits fraud. Industry experts anticipate these issues will be resolved over time, similar to how early e-commerce regulations evolved to protect both merchants and consumers. ## Interactive Consumer Testing Platform To increase transparency, Visa launched a new website allowing consumers to interact directly with the 22 models tested in the study. This platform provides insights into how these agents perform across 15 separate shopping expeditions. It serves as a practical tool for understanding the current limitations and potential of AI-driven shopping. ## Sources covering this story - https://fortune.com/2026/11/05/ai-agents-shopping-ranking-best-agent/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._