How to Chat With Files Using AI: Every Format You Can Upload

To chat with files using AI, you upload something — a report, a spreadsheet, a photo of a whiteboard, a recording — and ask it questions in plain language instead of opening it and reading or scrolling manually. The answer comes back written from the file's actual content, ideally with a citation pointing to where it found it.

People searching for “ai file chat” are usually asking a narrower question than it sounds: not “can AI read a PDF” (most tools can), but “which formats actually work, and does chatting with a spreadsheet or an image behave the same way as chatting with a PDF?” It mostly does — every format ends up as the same kind of indexed, searchable text, which is the idea behind document chat in general. This guide walks through the specifics format by format.

• PDFs — the most common case; see how to chat with a PDF.
• Word, Excel, and PowerPoint files — converted to text, and slide images get AI-captioned too.
• Web pages— paste a URL and it's fetched and indexed directly, no download step needed.
• YouTube videos — chat with the transcript instead of scrubbing the timeline.
• Images and scanned documents — AI-captioned or transcribed first, then indexed like any other text.
• Audio recordings — transcribed, then indexed like any other text.
• Ebooks — EPUB files work the same way as any other document.

A raw CSV export or a data file behaves a little differently from a formatted Excel workbook, but it's still something you can chat with: rows and columns get converted to text that keeps values tied to their headers, so you can ask “which row has the highest total in March?” instead of opening the file and scrolling or filtering by hand. HTML exports and zipped archives of files are handled the same way — each file inside gets processed on its own terms. The practical upshot is that “chat with files” isn't limited to documents that already look like reports; a plain data export you'd normally only open in a spreadsheet app works too, which matters if your source material is exported logs, survey results, or a table dumped out of another system.

The real advantage of chatting with files this way shows up once you stop treating each format as a separate tool. A single project can hold a PDF report, the spreadsheet it was built from, a slide deck summarizing it, and a recorded walkthrough — and one question can pull from all of them together, because retrieval doesn't care what format a passage originally came from. See what hybrid search is for why an exact figure in a spreadsheet still gets found even when the surrounding text doesn't look semantically similar to your question.

Doxy lets you chat with filesin whatever format they already exist in — PDFs, Office files, images, audio, web links, YouTube videos, and more — without converting anything by hand first. Add a few to a project and ask one question across all of them; every answer cites the exact source it came from. If you're still deciding between tools, see how Doxy compares as a NotebookLM alternative.