A friend’s daughter, halfway through a history degree, once spent four hours on a 600-word seminar post. Not on the research. On the sentence that opened it. She now drafts that sentence in ninety seconds with a prompt and spends the saved three hours reading the two sources she would otherwise have skimmed. That trade, not the app itself, is what sensible AI use looks like for a student.
Most lists of AI tools for students read like app store rankings: forty entries, no judgement, no prices. This one is narrower. It covers the handful of tools that earn a place in a real term-time week, roughly what they cost, and the places where they will quietly cost you marks if you let them.
Match the tool to the chore, not the hype
The students who get real value out of AI are boring about it. One tool for reading dense PDFs. One for editing prose. One for code. They don’t chase every launch, and they don’t pay for three subscriptions that do the same thing.
Everyone else installs six apps in freshers’ week and opens two by November. Before you download anything, write down the three tasks that eat your week. For most people it’s a variation on: reading more than you can absorb, writing more than you want to, and revising material you barely understood the first time. Those three problems need different answers.
Writing: structure and editing, not opinions
AI is a mediocre essayist and a very good line editor. Hand it your argument and it will point out where the logic sags. Ask it to invent the argument and you get 800 words of confident nothing.
Getting past the blank page
Outline generation is the highest-return use of a writing assistant in week one of a project. Describe your question, your word count and the three sources you’ve found, then ask for two competing structures. You’ll reject both and build a better third, which is exactly the point. If you want a sense of what a dedicated writing assistant adds over a general chatbot, this HyperWrite review covers its drafting and rewriting workflow in detail.
The last twenty minutes before you submit
Finish with a pass that catches the tedious errors: comma splices, four sentences in a row starting with “The”, passive drift, inconsistent citation style. Grammarly’s free tier handles most of that, and it runs inside your browser so you don’t have to think about it. Worth reading what Grammarly AI does well and where it falls short before you pay for Premium, because it will flatten your voice if you accept every suggestion. Take about half.
Revision that beats re-reading your notes
Re-reading is the most popular study method on earth and one of the least effective. It feels like learning because the words look familiar. What actually works is retrieval: testing yourself before you feel ready.
This is where AI earns its subscription. Upload your lecture slides to NotebookLM (free) and ask for fifteen exam-style questions built only from those sources. Answer them on paper, then ask the model to mark you harshly and name the specific idea you got wrong. Ten minutes of that beats an hour of highlighting.
For anything conceptually muddy, use the staircase trick: ask for the explanation a twelve-year-old would understand, then the version a lecturer would accept, then the gap between the two. That gap is what you don’t know yet. It’s roughly the AI-assisted learning framework I use for picking up unfamiliar topics, and it works just as well on a second-year statistics module as on a new job.
Coding coursework without skipping the learning
If you’re on a computing or data-heavy course, autocomplete is already the default. Verified students get Copilot free through the GitHub Student Developer Pack, and most undergrads use it whether or not their department has a policy on it.
The trap is subtle. Copilot is excellent at boilerplate (file reading, argument parsing, that one regex you’d spend forty minutes on) and weak at exactly the thing you’re being graded on, which is decomposing the problem. Let it autocomplete; don’t let it design. A workable test: if you can’t explain a line to your lab partner without reading it back, delete it and type it yourself. For a clear-eyed take on the trade-offs, read this 2025 breakdown of GitHub Copilot’s pricing and limitations.
Presentations, posters and group project chaos
Group work generates admin, and admin is where AI saves the most thankless hours. Otter.ai’s free plan transcribes 300 minutes a month, which is enough to stop someone being the unofficial note-taker every week. Gamma turns a rough outline into slides that don’t look like they were made at 2am.
Video is newer territory. A 60-second explainer for a module, a society or a placement application used to mean an afternoon in an editing suite. This InVideo AI review follows a full prompt-to-finished-video build, and the honest verdict is that it’s fine for talking-head explainers and poor for anything needing real footage. Know which one you’re making.
What the free tiers actually give you
- ChatGPT, Claude, Gemini: free tiers handle coursework comfortably if you keep prompts focused. Paid plans sit around $20 a month and mostly buy longer context and higher limits, not better answers.
- NotebookLM: free, source-grounded, and the single best hour of revision value on this list.
- Grammarly: free covers grammar. Premium is worth it only if you write every week.
- GitHub Copilot: free for verified students, otherwise $10 a month.
- Otter.ai: 300 free transcription minutes a month, which resets rather than rolls over.
- Perplexity or Consensus: useful free tiers for finding papers, but open the original every time.
Where these tools quietly cost you marks
Fabricated references are the big one. Models will invent a plausible journal article with a real-looking author and a DOI that goes nowhere. If you wouldn’t write a citation from memory, don’t accept one from a chatbot. Search tools like Consensus reduce the risk; they don’t remove it.
Voice is the second problem. Marker fatigue is real, and a marker who has read sixty essays that all open with the same three rhetorical moves is not on your side. The third is the reading you skipped. When AI summarises a paper, the summariser had the intellectual experience, not you, and it shows up in seminars long before it shows up in marks.
The twenty-minute Friday routine that makes it stick
Tools rot without a routine. Friday afternoon, before the week’s notes go cold:
- Dump each module’s slides and readings into its own notebook, so sources never blur together.
- Generate ten recall questions per module and answer three of them out loud.
- Write one paragraph in your own words on the idea you understood least this week.
- Delete any tool you didn’t open since Monday.
That last line does more work than the rest combined. A student with two well-used tools and a Friday habit will outperform one with fifteen subscriptions and no system, every single term. Run it for three weeks and you’ll know exactly which of these deserve a place in your second year.