Most articles about AI for college students are about whether students should use it. That ship sailed. The honest question is which tools are worth using, what each one is actually good at, and how to use them without quietly draining the learning out of a course you are paying for. This is a practical guide to the four free tools that have become the default for college students who use AI for studying — what each does well, what each fails at, and the hardest skill, which is knowing when to put the tool down and do the work.
None of this is about whether to use AI. It is about how to use it well.
ChatGPT (free tier) — what it is good at, what it is not
ChatGPT is the default. Originally released in November 2022 by OpenAI, it is the tool most college students try first, and for many it remains the only tool they use. The free tier is genuinely useful — most students never need to upgrade for studying purposes — but it is not the strongest tool for every job, and a serious student who only ever uses ChatGPT is leaving real capability on the table.
What it is good at. Conversational explanation of concepts you find confusing in your coursework. Drafting, rewriting, and tightening your own writing. Working through problem-solving steps when you are stuck and need a tutor at midnight. Generating practice questions on a topic so you can self-test before an exam. Translating between technical and plain-language phrasing — useful when a textbook is dense and you want a second pass.
What it is not good at. Citing sources reliably. ChatGPT can produce plausible-sounding citations that do not exist; for any work where the citation matters, ChatGPT alone is the wrong tool. Working with long-form documents you upload (the free tier's file-handling is limited; if you need to query a 200-page PDF, NotebookLM is a stronger fit). Real-time information — ChatGPT's free tier does not consistently search the web, and for time-sensitive questions you want a tool with explicit search built in.
How to use it well. Treat it as a smart, fast, occasionally-wrong study partner. Ask it specific questions, not vague ones. "Explain how Bayesian updating differs from frequentist inference, with a concrete example I might see in an introductory statistics course" beats "explain Bayes." Always check it on anything load-bearing — if the answer is going into your essay or exam, verify the claims against your textbook or another source. The tool is at its best as a starting point for your own thinking, and at its worst when it is doing your thinking for you.
Claude (free tier) — what it is good at, what it is not
Claude is Anthropic's chat product, available at claude.com, and has become the second-most-used AI tool for college students. The free plan includes chat on web, iOS, Android, and desktop, with code generation and data visualisation, content writing and editing, and text and image analysis. Anthropic does not publish exact daily message limits on the pricing page; usage limits apply, and Pro users get more usage plus access to Claude Code.
What it is good at. Long-form reading and analysis. Claude tends to handle long context windows comfortably — useful when you want to paste a chapter, a research paper, or a long set of class notes and have a conversation about it. Writing and editing tasks, particularly when you want a careful, considered response rather than a fast one. Code review and explanation, especially for students learning to read existing codebases.
What it is not good at. Real-time information without web-search activated; same limitation as ChatGPT free tier in this regard. Image generation. Heavy mathematical computation — for symbolic math or large numerical work, a calculator or specialised tool still beats both Claude and ChatGPT.
How to use it well. Reach for Claude when the task is reading-heavy. If you are working through a research paper for a literature review, Claude is often the better partner. If you are debating between two possible interpretations of a primary source, Claude's responses tend to weigh trade-offs explicitly rather than picking one and moving on. Pair it with ChatGPT — the two together cover more ground than either alone.
NotebookLM — for synthesis and research
NotebookLM is the free tool most students do not know about, and the one that most directly replaces a piece of work that used to take hours. Developed by Google Labs and first introduced in May 2023 as Project Tailwind, it became generally available in 2024 and now runs on Google's Gemini models.
The mechanism is different from ChatGPT or Claude. You upload sources — PDFs, Google Docs, websites, Google Slides, YouTube videos via transcripts, and text files — and the tool answers your questions strictly from those sources, with citations back to the specific passage in each source. It is a synthesis-and-search tool over your own document collection, not a general chat tool.
What it is good at. Working through a stack of readings for a course. You upload the syllabus' assigned papers, ask questions across them, and get answers with citations to specific paragraphs. Building study guides from your own course materials. Generating Audio Overviews — released September 2024 — that turn a notebook into a podcast-style two-host conversation, useful for review on the go. Comparing how multiple sources treat the same concept.
What it is not good at. Open-ended questions outside the scope of the sources you have uploaded. NotebookLM is deliberately constrained to the documents you give it; if the answer is not in those documents, it will tell you it does not know. This is a feature for academic work and a limitation for general study. It also will not help with current events or anything time-sensitive that is not in your uploaded sources.
How to use it well. Treat NotebookLM as the place where your course readings live. Upload everything assigned for a course at the start of term. As you go through the material, query the notebook to remind yourself of what each paper said about a topic, to find the source that supports a specific claim, or to compare frameworks across readings. The Audio Overview feature is genuinely useful for commute-time review of dense material — turn a notebook of papers into a thirty-minute podcast and listen on the way to class.
Perplexity — for sourcing and citations
Perplexity solves the citation problem ChatGPT and Claude cannot. Founded in August 2022 by Aravind Srinivas, Denis Yarats, Johnny Ho, and Andy Konwinski, Perplexity is built around real-time web search with source citations, so every answer comes with the URLs the answer was drawn from. The free tier offers basic search functionality with citations; Pro adds stronger security, document search alongside web content, and API access.
What it is good at. Research questions where you need to know not just the answer but where the answer came from. Fact-checking specific claims. Finding the most recent published source on a topic. Comparing how multiple sources describe the same event or finding. Discovering papers, articles, and primary sources you would not have found through a Google search alone.
What it is not good at. Long-form generation — Perplexity is built for search and citation, not for drafting essays or writing papers. The free tier limits the depth of follow-up before some features prompt you to upgrade.
How to use it well. Perplexity is the tool you use when ChatGPT or Claude gives you a claim and you want to verify it. Paste the claim, ask Perplexity for the source, and see what comes back. Use it for the early-research phase of any paper or assignment — Perplexity finds relevant sources, you read them, and the actual writing happens in your own document with the sources properly cited. Treat it as a faster, citation-aware Google.
How to use these without cheating yourself out of the learning
This is the section that matters more than the tool list above.
The honest version of "AI for studying" is that the tools are powerful enough to write the essay, solve the problem set, and produce the lab report — well enough to pass the course, sometimes well enough to get an A. The students who use them this way leave the course having paid the tuition and learned almost nothing the course was meant to teach. The credential on the transcript is real. The skill it is meant to certify is missing.
The version that does not cheat you involves a specific discipline: do the thinking first, use the tool to check or extend the thinking afterward.
For an essay, write your own outline before you ask any tool to help structure one. Draft your own thesis sentence before you ask any tool to refine it. Write your first paragraph before you paste anything into Claude. The tool then sharpens what you have written; it does not write what you have not yet thought.
For a problem set, work through the problem yourself first. If you are stuck, write down where you are stuck and what you have tried. Then bring the stuck-point to the tool and ask for the smallest possible nudge — a hint at the next step, not the full solution. The discipline is to use AI the way you would use a TA in office hours: to unblock specific moments, not to replace the work.
For a research paper, do the reading yourself. Use NotebookLM to navigate sources you have already engaged with, not to substitute for engaging with them. Use Perplexity to verify and extend, not to outsource the literature review.
The signal of whether you are using the tools well or being used by them is simple: ask yourself, after a week, whether you could explain the material in the course to someone else without the tool open. If yes, the tools are working for you. If no, you are accumulating a credential without the skill, and you will notice the gap when you reach the first job and the tool stops being the answer to every question.
When to switch from AI study to real practice
AI tools are excellent for studying — for understanding material, navigating readings, drafting and refining your own thinking. They are weaker as a path from understanding to capability.
The thing AI tools cannot give you is reps with feedback on real work in your target field. ChatGPT can explain how a market-entry analysis is structured. It cannot give you the experience of actually scoping one against a real brief, shipping it, having a practitioner review it, and revising in response. The first kind of learning makes you knowledgeable. The second kind makes you employable.
The transition from study to practice is what platforms like Ewance exist for. Challenge-based learning takes the topics you have studied and gives you a real brief to work them on. The deliverable you produce is reviewed against rubric criteria by someone who knows the field, and the feedback you get back is the kind your AI tools cannot generate — feedback grounded in what the work would need to be, in practice, to land in front of a hiring team.
Use the AI tools to study well. Then, at the point where studying alone stops moving you forward, switch to producing work that gets reviewed. The combination is what builds career-ready skill — neither the tools alone nor the practice alone gets you there.
If you want to see what the practice side looks like, see your options, or try a challenge for free and ship one piece end-to-end. The studying continues; the work begins.
