What can the assistant read?

The assistant reads the content that defines your business: website and subpages, the shop catalog with products, variants, prices and availability, opening hours and locations, documents and PDFs, FAQ and help articles as well as an internal knowledge base. Through a connection set up in the Custom package, also free times from your calendar and, with your inventory system connected, order or delivery status.

This way it answers not in general terms but specifically about your offering. How these sources are connected and kept up to date is described on the knowledge base and training page. For retail, catalog data is especially relevant, which the shop assistant evaluates.

For websites, clear limits apply. One run reads at most 25 pages, respects robots.txt, stays on the same domain and does not execute JavaScript; images, videos and other files are sorted out before fetching. A PDF therefore does not arrive through the page fetch but as its own source, and a larger website is split into several sources instead of being captured in a single run.

What cannot be read is rejected rather than guessed: a scanned PDF without a text layer delivers no content and reports that as an error instead of creating empty passages. If a run yields zero passages, the new version is discarded and the old state keeps answering. Which sources you can connect and how a run proceeds is described in the documentation.