Research the web visually by browsing pages
A commenter uses ChatGPT Agent Mode for visual web research, with the agent opening and working through pages.
What they did
The commenter says they used ChatGPT Agent Mode as a visual browser rather than a text crawler. It spun up a virtual computer, opened sites in a browser, looked at the rendered pages the way a person would, and clicked through live buttons and interfaces; they used it to search many online stores for a specific item and check whether each would ship to their country.
How it went
They say this worked much better on dynamic interfaces than text-based crawling and let the agent visit hundreds of stores to find what they considered the best option. The thread gives no hard metrics, exact prompt, or details on where they had to intervene.
Worth knowing
The main catch is that the source only gives a high-level description, so if you want to copy it you would need to supply the shopping goal and country constraints yourself.
Labels
they used it to search many online stores for a specific item and check whether each would ship to their country.
they used it to search many online stores for a specific item and check whether each would ship to their country.
they used it to search many online stores for a specific item and check whether each would ship to their country.
It spun up a virtual computer, opened sites in a browser, looked at the rendered pages the way a person would, and clicked through live buttons and interfaces
The commenter says they used ChatGPT Agent Mode as a visual browser rather than a text crawler.
The “Material” label failed its quality check and is not used for selection.
Details
- Agents
- ChatGPT Agent
- Source
- Source date
- Jul 10, 2026
- Accounts
- 1
- In registry since
- Oct 12, 2026