If you want to use AI to make notes, the best free option right now is Google's NotebookLM. You can upload documents, analyze them and produce summaries without it making anything up. I'll explain step by step which tool to use for each format and how to get all this information without spending a penny.
You can easily spend hours reading technical documents just to pull out a few key ideas or put together a basic outline. That can add up to two or three hours a day on repetitive tasks. I'll show you my simple system for automating this work directly with your own files.
What's the best AI for making notes?
If you search online, you'll probably end up putting your text into ChatGPT. It works quite well, but it has drawbacks. ChatGPT's free plan has an upload limit of up to 3 files every 24 hours and a maximum size of 512 MB per file. It also sometimes mixes what you upload with what it already knows from the internet.
In my view, the best free tool for making structured notes right now is NotebookLM. It's made by Google and designed for exactly this purpose. It creates virtual study notebooks where you upload your sources. It works with PDFs, Google Docs, text files, audio and even direct web links.
The most important thing about this tool is that it doesn't mix data. It uses a technology called source grounding, meaning it generates answers based on your sources. It only responds using what's in the documents you've provided. And, as I said, it's completely free. The standard plan lets you upload up to 50 documentary sources per notebook, with a limit of 500,000 words or 200 MB per file. That's plenty of space for a student or professional.
Another good option is the free version of Claude. It has a feature called Artifacts: a window alongside the chat that renders structured text or diagrams as you talk to it. If you want a full list of options, take a look at the best free AI tools on the market, but for making notes, I'd stick with Google's tool.
How to make notes from a PDF with AI, step by step
I'll show you the exact workflow for processing a document in NotebookLM. Imagine you have a technical report and need to extract the main points quickly.
The process is simple:
- Go to the NotebookLM website and create a new notebook.
- You'll immediately be asked to upload your information. Choose the PDF file from your computer or Google Drive.
- Once it's uploaded, you'll see an automatic tool called Study Guide in the sidebar.
- Click it and the AI will automatically generate a summary and a series of key questions about the text.
That's it. You have your basic outline. But there's more. NotebookLM has a feature called Audio Overview. It automatically generates a summary in the form of a conversational podcast between two AI voices. It takes your notes and creates a discussion in which two people explain the topic. It's useful for reviewing material while you're doing something else.
What if the PDF is protected or scanned?
Here's where you can run into real problems. There are two situations where this system won't work.
The first is a PDF protected against copying by a password or a DRM system. If it has this protection, the import into NotebookLM will fail automatically. The tool will tell you it can't read the file, and there's not much you can do except remove the password first, if you have permission to do so.
The second is a scanned PDF. If you receive a document made up entirely of scanned pages—that is, images of paper without an OCR text layer underneath—neither ChatGPT nor NotebookLM will be able to read it properly. Standard text tools can't extract words from an image without first passing it through optical character recognition software. So make sure you can select the text with your mouse before uploading it.
My method for using AI to summarize without errors
If you want to go beyond an automatic summary and ask the AI for something specific, you need clear instructions. If you're not sure how to structure them, I recommend reading what a prompt is first to understand the basics.
I use AI to extract information in very specific formats. A good prompt can make a big difference to the result. Here are two I use that you can copy and paste into ChatGPT, Claude or NotebookLM itself.
To create an executive summary of a long document, use this text: "Act as a university professor. Analyze the attached PDF and provide: 1) The central thesis in one paragraph, 2) Five key ideas explained in concise bullet points, 3) A glossary explaining the 5 most relevant technical concepts in simple terms."
If you want to study, the Cornell method works well. You can ask: "Extract notes from this PDF using the Cornell method, with three separate sections: 'Main notes' (essential ideas and formulas), 'Cues and review questions' (in the left-hand margin for self-assessment) and 'Final summary' (a 3-line overview)."
How do you stop AI from inventing information from a document?
One of the biggest concerns when using AI for study or work is that it will hallucinate—in other words, make up facts or dates that don't exist.
To avoid this with ChatGPT or Claude, the trick is in the prompt. Always add an explicit negative constraint at the end of your message. Something like: "Use only the information contained in the attached file. If the answer does not appear literally or directly in the text, respond with exactly: 'This information is not in the document'."
With NotebookLM, the problem is much smaller because it already works this way by default. Even so, the way to check is to look at the citation pop-up. At the end of each sentence Google generates, you'll see a number. Click it and the tool highlights the exact source paragraph in the PDF that supports that information. In other words, it takes you back to the original document so you can check for yourself that it hasn't made anything up.
Which AI tools can you use to make mind maps?
If you're more visual, you might prefer to see everything in a concept map. There are plenty of paid platforms for this, but I'll give you the free options that work from plain text.
| Tool | Input format | Free plan features | Cost |
|---|---|---|---|
| Markmap | Markdown (#, ##, -) | Open source | Free |
| Mermaid Live | Native mindmap syntax | No registration | Free |
| Whimsical | Visually assisted | 10 AI credits and 50 objects per board per month | Free |
Whimsical offers a free plan for building visually assisted mind maps, but you're limited to 10 AI credits and a maximum of 50 objects per board per month. For a large map, you'll reach that limit quickly.
In my view, the best way to generate mind maps automatically, without cost or watermarks, is to ask ChatGPT or Claude for the structure in text and then visualize it on free websites. If you're interested in tools like these without paying for subscriptions, you'll find a useful list at free AI tools.
Steps to create an interactive mind map for free
Let's look at how to build a map step by step using Markdown, a simple markup language, and the open-source tool Markmap. It's a free web converter that turns text outlines into interactive, downloadable mind maps in your browser.
- Go to ChatGPT or Claude, upload your document and enter this prompt: "Convert this text into a hierarchical mind map outline in plain Markdown: use
#for the central idea,##for main topics and hyphen lists-for subtopics. Do not include paragraphs or introductions." - The AI will return a structured code block. Copy it.
- Go to markmap.js.org and paste the text into the left-hand side.
- Your mind map will appear on the right. You can zoom in and out and download it to your computer.
If you prefer another format, you can use Mermaid.js. This tool supports a native syntax called mindmap, based on simple hierarchical indentation. The prompt is much the same: "Generate the structure of this document as a concept map in Mermaid.js syntax. Start with only the keyword mindmap, use two spaces of indentation per hierarchical level and use parentheses for second-level nodes. Provide only the code block."
Then go to the Mermaid Live Editor website and paste in the code. You can export the map as a vector SVG or a PNG image without registering or paying anything.
Can you summarize an entire book in one go?
A lot of people try uploading a very long PDF and asking AI for a complete summary in one go. This often goes wrong.
Language models have a context window—in other words, short-term memory. When they process very long documents, the context window becomes fragmented and they start leaving out important details from the middle of the book.
The official recommendation, and what I always do, is to ask questions by section or chapter rather than requesting a massive analysis of the whole document in one prompt. If you have a long document, ask for a summary by posing questions about each section or chapter individually. It takes a little longer, but the quality of the result is much better. And if you have to do this every day with lots of files, it may be worth looking into how to automate processes with AI using tools that split the document up for you.
In summary
Making notes, summaries and mind maps with AI doesn't require expensive subscriptions or complicated software. NotebookLM gives you a secure environment for analyzing PDFs without inventing data, and with Markmap or Mermaid you can create interactive diagrams by copying and pasting text. It's one of those improvements that saves you hours of the most tedious work. If you get stuck with any of these tools or want to learn how to set them up step by step for your business, you can take a look at the resources from El Atajo. Bye.


