Click add-to-cart button on a website with computer use
Btw, I tried to do the same with Gemini's computer model, and it couldn't find the add-to-cart button. Task: The user attempted to automate clicking an 'add-to-cart' button on a website using AI computer-use agents. Gemini's model failed to locate the button initially and could not click it after finding it, while Claude Computer Use succeeded on the first attempt. The source does not specify the exact website, product, or human oversight steps taken during the execution.; tools: Claude computer use, Gemini. Human role: The original post does not say how a person stayed involved.; limitations: The source is a brief social media post lacking technical details on the website structure, browser environment, or specific prompts used. It does not define safety protocols, cost implications, or integration requirements for production use. Reliability is limited to a single anecdotal comparison without controlled testing conditions..
Tools
Published on Agent Work Atlas:
Overview
Btw, I tried to do the same with Gemini's computer model, and it couldn't find the add-to-cart button.
Workflow
The user attempted to automate clicking an 'add-to-cart' button on a website using AI computer-use agents. Gemini's model failed to locate the button initially and could not click it after finding it, while Claude Computer Use succeeded on the first attempt. The source does not specify the exact website, product, or human oversight steps taken during the execution.
Reported result
Claude Computer Use successfully clicked the add-to-cart button on the first try, whereas Gemini failed to complete the task. No quantitative metrics such as time saved, error rates, or business impact were reported. The outcome is based solely on the author's qualitative observation of agent performance.
Human role
The original post does not say how a person stayed involved.
Limitations
The source is a brief social media post lacking technical details on the website structure, browser environment, or specific prompts used. It does not define safety protocols, cost implications, or integration requirements for production use. Reliability is limited to a single anecdotal comparison without controlled testing conditions.
Try it
Act as a QA engineer testing AI computer-use agents on a specific e-commerce workflow. I will provide the target URL [replaceable input] and the product page path [replaceable input]. Your task is to compare the performance of Claude Computer Use and Gemini on the 'add-to-cart' action. First, perform a dry run on a sample product to verify button detection and clickability. Show me the screenshots or logs of the attempt before proceeding. If the agent fails to find the button, report the error and stop. Do not execute any real purchases or submit forms without my explicit confirmation. Measure success by whether the cart count increments correctly. If inputs are missing, ask for the specific URL and product details.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.