I Sent 12 AI Shoppers Through a Broken Checkout, but Only One Bought Anything
I tested an e-commerce checkout with 12 AI shoppers, found a pricing bug, patched it, and reran the tests. Here's what changed and what still failed. TL;DR Th
An AI system sent 12 simulated shoppers through an e-commerce checkout process, but only one made a purchase. The store was selling three telescopes for $101.96 each, plus $26.97 shipping, totaling $128.93. The AI shoppers were instructed to verify the quantity, price, shipping, and total before proceeding. The developers used Playwright to control the browsers and Muse Spark to simulate actions like clicking, selecting, typing, pressing keys, scrolling, and giving up.
The developers built different shopper profiles with various browser conditions, such as slower CPU and network settings, enlarged text, and accessibility features. The checkout script placed the order, but the AI shoppers were tasked with checking the cart total for accuracy. Ten shoppers complained about the same total, and one shopper successfully purchased the telescopes before a patch was applied.
The other eleven shoppers left due to issues with the quantity calculation. The developers identified and fixed bugs related to the checkout process, including an issue with the native quantity dropdown and a problem with the screenshot not showing the open menu. After fixing the bugs, nine out of twelve shoppers were able to successfully purchase the telescopes.
Written by urgent.news from HackerNoon's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.