There are dozens of AI tools. You do not need dozens. You need five, and you need to know which one to reach for in which situation. Pick the wrong tool and you waste twenty minutes. Pick the right one and you get a better answer than you would have gotten by yourself in two hours.
Here is the short list, what each one is actually good at, and the one weakness you should never forget.
ChatGPT
What it's for: General-purpose writing, brainstorming, explanations, code, light research. The Swiss Army knife. If you only use one tool, this is the one most people start with.
Killer student use case: Drafting and refining essays, lab reports, and emails. ChatGPT's prose is competent and easy to steer. It is also genuinely good at acting as a Socratic tutor if you tell it to.
Act as a Socratic tutor for organic chemistry. I'm trying to understand
SN1 vs SN2 reactions. Don't give me the answer. Ask me questions one at
a time that lead me to figure it out. Start with the simplest one.
Weakness: It fabricates citations confidently and frequently. Never let ChatGPT give you a quote, a study, or a source URL and just trust it. If it cites a paper, you must open the paper. About a third of the time it does not exist.
Claude
What it's for: Long documents, careful writing, code review, anything where you want a model that reads like an actual thoughtful human being and not a corporate brochure.
Killer student use case: Feeding it an entire essay or a full chapter PDF and asking for structural feedback. Claude has a massive context window and is unusually good at sustained, nuanced critique. It also tends to push back on bad arguments instead of flattering you.
Here's my 2,500-word essay on the causes of WWI. I'm arguing economic
factors mattered more than nationalism. Tell me the three weakest parts
of this argument and what a strong critic would say. Be direct.
Weakness: Sometimes it refuses things it shouldn't, or hedges where you wanted a confident answer. Tell it explicitly what you want: "Give me a direct answer, not a list of considerations."
If you want a focused intro, see Micro: Claude for Beginners and Prompt Engineering with Claude.
NotebookLM
What it's for: Turning your own course materials into a queryable knowledge base. You upload syllabi, lecture slides, PDFs, your own notes — up to 50 sources — and ask questions across all of them.
Killer student use case: Midterm prep. Dump every lecture slide deck and reading from the semester into one notebook. Ask "what topics has the professor emphasized most across all lectures?" or "what's the connection between Lecture 3 and Lecture 8?" The answers cite the exact source page.
The Audio Overview feature also generates a 10-minute podcast from your sources. It is shockingly good for review while you're walking to class.
Weakness: It will only use the sources you give it. Forget to upload Lecture 7 and it will confidently answer questions as if Lecture 7 doesn't exist. Garbage in, garbage out — but at least the citations are real.
Perplexity
What it's for: Research with real, clickable sources. Think of it as Google + ChatGPT in one box. Every claim links to a webpage you can verify.
Killer student use case: Starting a research paper. Most students start with a vague Google search, click around for an hour, and produce a Word doc full of unsourced assertions. Perplexity gives you ten cited sources in thirty seconds and you can drill into any of them.
What are the strongest empirical critiques of the broken windows theory
of policing published since 2015? Prioritize peer-reviewed studies.
Weakness: It's only as good as the web. For deep academic work, follow up by searching Google Scholar for the citations Perplexity surfaces. Treat it as a starting point, not the destination. Go deeper with Perplexity AI for Research and Micro: AI Research.
Gemini
What it's for: Anything inside Google's ecosystem. Live web access. Multimodal tasks involving images, video, and Google Drive integration.
Killer student use case: "Here's a photo of the equation my professor wrote on the board. Solve it and explain each step." Gemini's image understanding is excellent. It also natively reads files from your Google Drive, so you can ask it questions about a doc you wrote last semester without uploading anything.
Weakness: It can be cautious to a fault and sometimes lectures you about its own limitations instead of just answering. It also hallucinates differently than ChatGPT — confidently inventing specific dates and quantities. Verify numbers.
A simple rule for choosing
You don't need a flowchart. You need three reflexes:
- Need real sources? Perplexity. Always Perplexity first. Then verify on Google Scholar.
- Working with your own materials — slides, PDFs, your notes? NotebookLM. Nothing else comes close.
- Everything else — drafting, brainstorming, coding, tutoring, debugging, casual questions — pick ChatGPT or Claude based on the task. Claude for long-form, careful work and writing critique. ChatGPT for fast iteration and general-purpose chat. Gemini if you live inside Google Workspace.
That's the whole decision tree.
What you can ignore for now
Midjourney, Stable Diffusion, Runway, ElevenLabs, Suno — image, video, and voice tools. They are fun and you should play with them, but they are not core to studying. Don't let the shiny stuff distract you from the workhorses.
Same with the dozens of "AI study app" startups. Most of them are thin wrappers over ChatGPT charging you $15/month for a worse interface than ChatGPT's free tier. Use the underlying model directly until you have a specific reason not to.
One more thing
Get used to switching mid-task. A serious workflow looks like: Perplexity to find sources, NotebookLM to query them in detail, Claude to draft an outline, ChatGPT to draft a section, then back to Claude to critique it. The students who treat AI tools as a single hammer will lose to the ones who keep five different ones on the desk.
Five tools. Pick the right one. Move on.