Verified purchaser
Couldnt Scrape a Simple Blog
I thought I’d try the AI builder by giving it a simple instruction: scrape a website’s blog titles and URLs for the first 10 pages of the blog. Then, I asked it to export the data to a CSV with columns for blog titles and URLs. It was cool to watch the live browser move around, but after trying the query five times, it couldn’t do what I wanted. It kept hallucinating and repeating scraping one page twice. It couldn’t even continue to pages 3, 4, or 5. In the end, I lost 15,000 credits even though the output was incorrect, and I received no refund for the wrong delivery. I’m not sure if I’ll wait for future updates or just ask for a refund. I expected it to work out of the box, but it didn’t.
Claire_BrowserAct
Oct 30, 2025I'm really sorry you had this frustrating experience. This isn't the result we want for any user, and I completely understand your disappointment.
What Happened:
You used the AI Agent (Beta) mode, which is more experimental and can have inconsistencies like:
Repeating actions (hallucinating)
Failing to navigate pagination correctly
Variable results between runs
This is exactly the kind of issue we're working to improve in Agent mode—but I understand that doesn't help you right now.
The Better Approach for Your Use Case:
For reliable, repeatable scraping (like blog titles and URLs across multiple pages), you should use Workflow mode instead of Agent mode:
Workflow mode advantages:
✅ Consistent, predictable results
✅ Reliable pagination handling
✅ No hallucinations or random behavior
✅ Perfect for production use
Agent mode (what you used):
⚠️ More experimental
⚠️ Can have inconsistent outputs
⚠️ Better for one-off exploratory tasks
About the Credit Usage:
I hear you on losing 15,000 credits for incorrect output. That's frustrating, especially when you're testing and trying to get it right.
Please reach out directly so we can help:
Discord: https://discord.com/invite/UpnCKd7GaU (open a ticket)
Email: [email protected] or [email protected]
Include:
Your account details
The workflow/task IDs
What happened
We want to make this right and help you succeed with BrowserAct.
Moving Forward:
If you want to try Workflow mode:
We'd be happy to help you build a proper workflow for scraping blog titles/URLs with pagination. This should work reliably and give you the results you expected.
If you prefer a refund:
Contact AppSumo or reach out to us directly. We understand if this experience has put you off, and we won't stand in your way.
What We're Learning:
Your feedback is important. The Agent mode needs clearer warnings about its limitations, and we need to better guide users toward Workflow mode for production tasks like yours.
Thank you for giving BrowserAct a try, and I'm sorry it didn't work out of the box as expected. Whatever you decide, we're here to help.
Claire & the BrowserAct Team