This video demonstrates the test environment used for our study. To ensure consistency across trials, all participants operate on a standardized Windows desktop device. The left side of the screen features a browser window displaying the CAPTCHA challenge. Participants have unlimited opportunities to interact with the CAPTCHA manually. They can inspect the browser console to verify if the challenge was successfully solved, switch between different CAPTCHA types via the URL bar, or restart the current challenge as needed.
On the right side, a separate window is dedicated to interacting with GPT-4o using both image and text inputs. Participants can use the snipping tool to capture and annotate portions of the screen, which they then incorporate into their prompts. Additionally, a live display in the bottom-right corner shows the current mouse coordinates. This allows users to compare the coordinate-based outputs generated by GPT-4o with the CAPTCHA locations on the left.