Visa test: AI shopping agents failed 42.8% of runs
Visa has released the first AI shopping benchmark, the Visa Agentic Commerce Benchmark, running 22 AI models through realistic shopping scenarios. In the first round, 42.8% of test runs failed; models from Google and Meta led, with Claude third (5 November 2026).
Who is affected and how
- fintech company
- The benchmark is a yardstick for demonstrating the reliability of agentic shopping features.
- banking professional
- With agents initiating card transactions, the high failure rate means banks should prepare for more disputes and refund cases.
- retail customer
- An order entrusted to a shopping agent is worth checking before the final step.
Sources
- Visa’s AI Shopping Agents Failed in More Than Four in 10 Testswww.tokenpost.com · Nov 5, 2026
- Visa’s AI Shopping Agents Failed in More Than Four in 10 Tests | TokenPosttokenpost.com · Nov 5, 2026
- Visa tested AI shopping agents—and they failed nearly 50% of the time | Fortunefortune.com · Nov 5, 2026
- Visa study finds AI shopping agents succeed only about 57% of the timecryptobriefing.com · Nov 5, 2026
- AI Agents Fail Nearly Half of Shopping Tasks in Visa Testforklog.com · Nov 6, 2026
- Visa tested AI shopping agents – here’s how they performed | Chain Store Agechainstoreage.com · Nov 6, 2026
- Visa launches benchmark for AI shopping agents | The Paypersthepaypers.com · Nov 6, 2026
- Visa Releases First-of-Its-Kind Industry Benchmark for AI Shopping; Almost Half of Test Runs Fail | Visausa.visa.com · Nov 5, 2026
Machine-generated summary based on the titles and excerpts of the source articles. The source article is the authority for the facts. News monitoring, not investment, financial or tax advice.