Upload a PDF and ask it questions directly, like texting someone who has already read the whole thing. Answers are grounded in your document's actual text, not generic AI knowledge, so it stays accurate to what's actually written.
The Summarizer gives you one fixed overview in a single pass. This is a real back-and-forth: ask a follow-up, get more specific, jump to a different part of the document, all in the same conversation without re-uploading anything.
Skimming a long contract or research paper and needing a quick answer to one specific question, "what's the notice period", "what was the total in section 3", without reading the whole thing end to end.
It's instructed to answer only from the document text and conversation so far, and to say plainly when something isn't in the document rather than making it up.
Only if there's already selectable text. Run OCR PDF first on a scan to give it a real text layer.
Yes, it keeps the last few exchanges as context, so follow-ups like "what about the second one" work naturally.
Very long documents get truncated to the first several thousand characters. For a huge document, ask about the section you actually care about.