How to chat with a PDF

A 60-page supplier contract arrives the day before a call, and the only questions that matter are when it renews, how much notice it takes to end it, and who carries the liability if a delivery goes wrong. Reading the whole thing is the careful option. Asking the document directly is quicker, as long as you can see where each answer came from.

This guide covers the usual ways people chat with a PDF, where those methods get awkward, and how to do it in Fabric, where the PDF stays in your library and the AI assistant answers with references back to the file.

How people usually chat with a PDF

The most common route is to upload the file to ChatGPT, Claude or Gemini and start asking. For one document in one sitting, that works well. The PDF lives inside that conversation, though, so a question next week means finding the old chat or uploading the file again, and comparing it with a second document means attaching both to the same thread.

Dedicated chat-with-PDF websites take the same approach with a narrower focus. They’re quick to try and often free for a few files. They also add one more place where documents end up, separate from wherever the rest of your files are kept.

Then there’s the manual method: Ctrl+F for keywords, a highlighter, and reading the sections that look relevant. It’s reliable for exact terms like a clause number. It struggles when the question is phrased differently from the document, such as asking how to get out of a contract when the PDF talks about termination for convenience.

For background on what happens when an AI tool reads a document, How LLMs use your files explains the mechanics.

Step 1: Add the PDF to Fabric

Drag the PDF into the Fabric web or desktop app, paste it, or click Upload a file. On a phone, open the PDF in whichever app it’s in, tap Share and choose Fabric; the mobile apps are available for iOS and Android. A PDF that came as an email attachment can be forwarded to your Fabric email address instead of being downloaded first.

Once it’s in, Fabric reads the text inside the PDF, writes a short summary, and creates embeddings so that search and the assistant can match passages by meaning. There’s no need to tag or file the PDF before you ask about it.

Step 2: Point the assistant at the right document

The AI assistant can be limited to a folder. Open a folder, open the assistant, and it automatically limits itself to what’s in that folder. For a single contract, a folder holding that PDF (plus any schedules or amendments) keeps the answers tied to the right paperwork.

You can skip opening the folder. Right-click it and add it to the assistant’s context, which has the same effect. That helps when you’re working on something else in Fabric and want to ask about a set of papers without leaving it.

Leave the assistant unscoped and it searches across all of Fabric and any apps you’ve connected. That suits questions like “which of my PDFs mentions the 2025 price increase?”, where you don’t yet know which file holds the answer.

Step 3: Ask questions and check the citations

Type the question the way you’d ask a colleague who has read the document. “What’s the notice period for ending this agreement?” and “Explain section 4 in plain English” both work. The assistant answers from the PDF and cites its sources with clickable references, so you can open the passage it used and read the original wording. There’s more on how this works on the document chat page.

Follow-up questions keep the thread of the conversation. After asking about the notice period, you can ask whether it changes after the first year without restating which contract you mean. For documents where exact wording carries weight, such as leases, policies or medical letters, treat the answer as a guide to the right passage and read that passage before acting on it.

You can also choose which model answers. One Fabric subscription covers models from the major labs, including Gemini, OpenAI, Claude, DeepSeek, Grok and Qwen, so a tricky question can be put to a second model for comparison.

Step 4: Keep the answers next to the PDF

Answers can be saved as notes, so a list of the contract’s key dates sits in your library beside the PDF it came from. You can also highlight and comment on the PDF itself, which is a good way to mark the clauses the assistant pointed you to.

When you need an exact phrase later, search reads the text inside PDFs and opens results at the exact page. Put the phrase in quotes for strict matching, which helps with reference numbers and defined terms.

The AI assistant is on the free plan with limited use, and heavier use needs a paid plan. The pricing page lists what each plan includes.

Frequently asked questions

Can I chat with a PDF for free?

Yes. The free plan includes file uploads and limited AI use, so you can add a PDF and ask the assistant about it. Heavier AI use needs a paid plan.

How do I chat with several PDFs at once?

Put them in one folder, open it, and open the assistant. It’s limited to that folder and analyzes everything in it, so you can compare documents in a single conversation.

Does the AI show where its answer came from?

Yes. Answers cite their sources with clickable references, so you can open the document and check the wording yourself.

Can I ask follow-up questions about the same PDF?

Yes. The assistant keeps the context of the conversation, so a follow-up can refer back to the previous answer without repeating the document name.

Which AI model does Fabric use to answer questions about a PDF?

You choose. One subscription covers models from Gemini, OpenAI, Claude, DeepSeek, Grok, Qwen and others.

Can the AI read a scanned PDF?

Fabric reads text inside PDFs and images, including photographed text, so scanned pages are searchable and available to the assistant. How to search your scanned documents covers this in more detail.

Do I have to upload the PDF again every time I want to ask about it?

No. The PDF stays in your Fabric library, so you can open the assistant on it next week or next year and pick up where you left off.

Can ChatGPT or Claude read the PDFs I keep in Fabric?

Yes. Connect them through the Fabric MCP server, and they can search Fabric and read file contents. Sign-in uses OAuth, so there are no API keys to manage.