What Visa Found
Visa announced its Agentic Commerce Benchmark on November 5, 2026. It put 22 AI models from 12 labs through 15 shopping challenges, such as buying a TV, a laptop or a personal care appliance. The overall pass rate was 57.2%, so 42.8% of runs failed. The best model passed 82.2% of attempts and the worst 28.9%.
Other findings: about 1 in 2 (54%) of failed runs still ended in the AI making a purchase, and all three attempts passed only 34.2% of the time across models and scenarios. The five models that uncovered the most buying factors averaged a 70.2% pass rate versus 44.4% for the five that uncovered the least. Visa plans to extend testing to checkout, payment and order placement. The report does not name the individual models.
Three Steps for Small Merchants
- Make your product details unambiguous. Visa found that models which uncovered more buying factors passed more often. Clear sizes, specs, prices and stock on each listing leave less for an agent to guess. This is our advice.
- Keep your return and support details easy to find. Because some failed runs still ended in a purchase, a customer may arrive having bought the wrong item. Clear policies and a quick way to reach you help. This is our view.
- Own a storefront that is accurate at the source. AI agents read what you publish. A website you control gives every shopper, human or AI, one correct version of your catalogue. This is our advice.
The benchmark covers product choice, not yet payment, and its results are a snapshot. Still, it shows why accurate, complete product pages matter as more shopping starts with an AI.
