ChatGPT, Claude, Gemini or NotebookLM: which suits your course files?
They're all good, and they all change every few months, so any ranking goes stale quickly. How each one handles your files changes less often, and it makes more difference to a study session than a benchmark score does.
Two ways of working with documents
Chat with a document
ChatGPT, Claude and Gemini are general chat assistants. You paste or attach your material in a conversation and ask about it. They're flexible. They'll explain a topic three different ways, quiz you, rewrite your notes, and fill in from general knowledge when your notes are thin.
Your material usually stays with that one conversation. Start a new chat and you start over, unless you use a feature that keeps files across chats. Most of these assistants have something like that now, often called projects. What each plan includes keeps changing, so check the help pages for yours.
A notebook of sources
NotebookLM, from Google, works differently. You make a notebook (one per course, say) and add sources to it: PDFs, Google Docs and Slides, pasted text and a few other types. Its answers come from those sources, with citations that take you to the passage each point came from.
It's built to stay inside your material. That's great when you want to know what the reading actually says. It's less of a general tutor, though. When your sources don't cover something, it tends to tell you so instead of filling in. NotebookLM also has built-in study features that turn your sources into guides and other formats. Those change a lot, so see what's there when you try it.
What to compare
Whichever ones you're looking at, these are the questions that matter for course files.
- Which files does it take? They all take PDFs and pasted text. Support for PowerPoint, Word and spreadsheets varies, and so does whether it reads the images inside a file.
- How much can you give it? Free plans limit uploads, file sizes and messages, and the numbers change. Your plan's help page has the current ones.
- Can you see where an answer came from? NotebookLM's citations do this by design. A chat assistant can quote its source if you ask, but you have to ask, and you have to check the quote is real.
- How well does it hold up over a long session? Every chatbot can only keep so much in mind at once, and with a big batch of notes the early material can slip. Paste or attach? shows how to check what it actually read.
- Does it stay inside your material? Mixing in general knowledge helps when you need something explained. It's a problem when your course defines something its own way.
- What happens to your files? Each service has its own data policy, and some use your conversations to improve their models unless you turn that off. Look in the privacy settings.
- What does your university already give you? A lot of universities provide an AI tool through your student account, sometimes with different data terms from the personal version. Your IT or library pages will say.
Matching the tool to the job
- For getting your head around a hard topic, a chat assistant is usually better. You can push back, ask for another example, or say "explain it as if I missed the lecture" and keep going until it makes sense.
- For finding what the readings say, a notebook of sources is the better fit. The citations take you straight to the passage, which is what you want when you're planning an essay or checking a definition.
- For revising a whole course, a notebook you add to every week keeps everything in one place for the exam. A chat assistant with projects can do something similar.
- For quizzing and flashcards, any of them will do. Chat assistants are good at one-question-at-a-time quizzing. See flashcards and practice exams for prompts.
- For calculations on a spreadsheet, use an assistant that can run code on the original file. Giving ChatGPT an Excel or CSV file explains why.
- For diagrams and scans, pick one that reads images, and upload the originals.
The comparison that tells you the most is one you run yourself. Give two of them the same notes and the same three questions about this week's material, then see which answers are more accurate, easier to check, and closer to how your course teaches it. Use that one for this course. Another course might come out differently.
One text file works with all of them
CourseFeed turns your slides, readings, Word documents and spreadsheets into one Markdown document, with a # Source: heading for each file. Markdown is plain text, so the same file works in any of these tools.
For ChatGPT, Claude or Gemini, press Copy for AI and paste if it's a lecture or two. For a bigger batch, press Download .md and attach the file. If an app won't take a .md attachment, rename it to .txt. Nothing inside changes.
In NotebookLM, add it as a pasted-text source or upload the file. The whole batch then counts as one source.
It also means that if you switch tools halfway through the term, you don't have to upload ten files all over again. The same document works in the next one.
There's one trade-off with NotebookLM. A combined file is a single source, so citations point into that one source instead of to a named file. The # Source: headings still show which file each passage came from. If NotebookLM already accepts your file types, uploading them separately can work better there, since you get one source per file and can switch individual sources on and off. The combined file is most useful for files it doesn't take directly, and for chat assistants with tight upload limits.
Either way, scanned pages have no text to convert, so upload those as they are to a tool that reads images.
True of all of them
- They can be wrong and sound completely sure. Check anything that'll be on the exam against the slides or reading.
- Naming the source helps. "Using the Week 5 reading" gets answers tied to your material instead of general knowledge.
- One chat or notebook per course keeps different courses from getting mixed up.
- A summary isn't studying. These tools help most when they make you answer questions, not when they hand you something to read.
Academic integrity
Rules about AI differ between schools, between courses, and even between assignments in the same course. Using AI to study your own course material is usually treated differently from using it to write work you hand in, but not always. Some courses restrict AI completely, and some don't let you upload lecture material (which belongs to your lecturer or university) to outside tools. Check the syllabus or ask your instructor. Where AI is allowed on an assignment, a lot of courses expect you to say how you used it, so keep notes as you go.