Add url: webpage upload is inventing urls for child pages

When adding an url, the next step is to select linked / child urls. The problem: most url are invented, they are following a wrong path (invented on the provided url)

Hey Ulrich, are you referring to invented pages in the knowledge base upload? That shouldn’t be the case, because that feature doesn’t actually use AI, it just uses web-scraping. Can you provide a screenshot?


Hi Mike, here is a screenshot of made up urls. In fact, all urls that follow the main url are made up. It’s reproductible for all urls that i added. I had to deselect all, and select only the first.

1 Like

Hi @ulrich,

Here’s how it works: we first check the sitemap of the website. If we can’t find it there, we retrieve it from the anchor links on the website that was input.