---
title: "Blog"
canonical: https://www.checkoutbench.ai/blog
---

# Blog

Research and notes from Checkout Bench on how AI shopping agents do at real store checkouts.

- [Checkout Bench in Action: See Amtrak from Search to Payment](https://www.checkoutbench.ai/blog/amtrak-checkout) (Research, November 18, 2026): See how two Pacific Surfliner tickets went from a station search to a $144 checkout, with all 34 screens shown side by side with our annotations and digital twins. ([markdown](https://www.checkoutbench.ai/blog/amtrak-checkout.md))
- [OpenAI's plugin guide asks stores to return the facts our browser agents get stuck on at checkout](https://www.checkoutbench.ai/blog/openai-plugin-guide-checkout-failures) (Research, November 18, 2026): We matched each field in OpenAI's commerce plugin guide to the step where agents stopped in our runs at real store checkouts. ([markdown](https://www.checkoutbench.ai/blog/openai-plugin-guide-checkout-failures.md))
- [Introducing Checkout Bench](https://www.checkoutbench.ai/blog/introducing-checkout-bench) (Company, November 17, 2026): We're an LA-based, AI-first data lab that builds bespoke training environments for top retailers. ([markdown](https://www.checkoutbench.ai/blog/introducing-checkout-bench.md))
- [Getting into a store is harder than buying from it](https://www.checkoutbench.ai/blog/getting-in-is-harder-than-buying) (Research, November 11, 2026): About half of agent sign-ins work, and once signed in at a big store, an agent rarely hits a bot wall at checkout. ([markdown](https://www.checkoutbench.ai/blog/getting-in-is-harder-than-buying.md))
- [Reading Visa's AI shopping benchmark next to 904 real checkouts](https://www.checkoutbench.ai/blog/visa-agentic-commerce-benchmark) (Research, November 10, 2026): Visa's benchmark and ours both lose about half their runs, and both lose them at the step where the agent has to work out exactly what the shopper wants. ([markdown](https://www.checkoutbench.ai/blog/visa-agentic-commerce-benchmark.md))
