Visa's AI Shopping Test Reveals Models Fail 43% of Runs as Agentic Commerce Faces Trust Hurdle
Visa launched its first AI Agentic Commerce Benchmark, testing 22 AI models from 12 labs across 15 shopping tasks. The results show AI shopping agents passed only 57.2% of test runs, with Google and Meta's models leading at 82% success rates. In 54% of failed attempts, AI agents still made purchases that didn't meet customer requirements, highlighting reliability concerns in agent-driven shopping.