Visa's Agentic Commerce Benchmark Finds 43% Failure Rate Across 22 AI Models

Visa's Agentic Commerce Benchmark tested 22 AI models from 12 labs on 15 shopping tasks, with an overall pass rate of 57.2%. Alarmingly, in over half of the failures, agents still completed purchases even when no matching product was found.

2026-11-10 ~ 2026-11-11 · 2 related posts

• Visa's agentic commerce benchmark: 57.2% pass rate, agents buy in 54% of failures — krishnan · 2026-11-10
• Visa tested 22 models on 15 shopping tasks: ~43% failure, agents bought wrong items anyway — Thick_Mulberry_2235 · 2026-11-11