AI Study Tools, Compared: What Each Is Actually Good At
AI Study Tools, Compared: What Each Is Actually Good At
I'm going to skip the part where I tell you AI is reshaping education and you need to find the right tools to keep up. You know.
Instead I want to do something most comparison posts don't: be honest about what each category of AI study tool is actually for. Some are excellent at narrow things and bad at others. None of them are bad products in general. The mistake students make isn't picking a "wrong" tool — it's picking a tool that's solving a different problem than the one they have.
I'll group these by category, not by brand, because the brand is often less important than the shape.
The four categories that matter
There are roughly four kinds of AI tools a student might use:
- Generic AI chatbots — ChatGPT, Claude, Gemini
- Specialized math/computation tools — Photomath, Wolfram Alpha, Symbolab
- Writing tools — Grammarly, Hemingway, the writing features inside Word/Google Docs
- Integrated study systems — Fennie, and a small number of similar products
Each is actually good at something. Each fails at something else. Let me go through them.
Generic AI chatbots
What they're good at: explaining a single concept on demand, brainstorming, drafting, summarizing a passage you paste in, walking through a problem when you ask the right way.
What they're bad at: knowing what you should be studying right now. Knowing what you've already learned. Knowing your professor's chapter ordering. Spaced repetition. Generating practice that's actually scoped to your course. Telling you "no, you should review chapter 4 first."
If you've used ChatGPT or Claude for school, you already know the shape. You ask, it answers. The conversation has no memory of last Tuesday. It doesn't know you have an exam Friday. It doesn't know that you got the same kind of mass-balance problem wrong three times last week.
For specific use cases, these tools are excellent. "Explain partial derivatives like I'm a calc 1 student" — Claude will do this beautifully. "Help me brainstorm a thesis for a paper on Reconstruction" — ChatGPT will give you eight angles and ask which one resonates.
For the actual job of getting through a 16-week term across four or five courses, they're not designed for it. Not because they're bad. Because that isn't what they are.
Who should use them
A self-directed student in a single course who knows what they need help with on a specific evening. A grad student doing focused reading. Anybody who already has a study system and just needs occasional explanation.
Who probably shouldn't lean on them
A pre-med taking five classes who needs structure across the term. A 1L for whom the workload itself is the main challenge. Anyone who's described their last semester as "I was always behind and didn't know what to do first."
Specialized math and computation tools
Photomath, Wolfram Alpha, Symbolab. These are the tools that point at a math problem and produce a step-by-step solution.
What they're good at: getting a problem solved. Showing the steps clearly. Handling integration, ODEs, matrix algebra, system solving. Wolfram in particular is unmatched for any computation that has a closed-form answer or a clean numeric one.
What they're bad at: teaching you to do the next problem yourself.
This is the most honest sentence I can write about these tools. They are tools for getting answers. They are tools for checking your work. They are reference tools. They are not study tools, in the sense that they don't make you better. If you use Photomath every time you get stuck, you'll still be stuck on the exam, because the exam doesn't have Photomath.
This isn't a moral problem. Wolfram is a calculator. A calculator is fine. The mistake is treating Wolfram as a tutor.
Where they fit in a real study workflow
Use them to check your work after you've solved a problem yourself. Use them when you've genuinely tried and need to see the step you missed. Use them to verify a numerical answer at the end of a derivation.
Don't use them as your first move. The cost is your retention.
Writing tools
Grammarly, Hemingway, ProWritingAid, the writing assistants now built into Word and Google Docs.
What they're good at: catching grammar errors, flagging awkward phrasing, suggesting clearer constructions, basic citation help.
What they're bad at: making your writing actually good.
A clean Grammarly score does not mean your essay is sharp. The best essays I've read had a fair number of intentional rule-breaks. Grammarly will flag those. The students who turn off the suggestions are usually the ones writing better.
Where these tools genuinely earn their keep: ESL students, students with dyslexia, anyone who consistently misses the same kind of small error. For those students, real-time writing feedback can be a quality-of-life win.
For everyone else, they're fine. Use them. Don't rely on them. They're not going to make a B paper into an A paper. The thinking does that, and the thinking happens before you start typing.
Integrated study systems
This is the category I work in, so I'll try to be precise rather than promotional.
An integrated study system is software that knows your courses, your calendar, what you've learned, what you've struggled with, and uses all of that to drive a daily plan and the way it tutors you.
The shape: you drop in a syllabus, the system maps the term. You add deadlines. Each morning you get a small specific plan — three or four items, usually 30 to 90 minutes total. The system has chat (that teaches rather than tells), notes (markdown plus LaTeX), generated quizzes from your notes, spaced-repetition flashcards, essay support. Underneath it all is memory: the part that tracks what you skipped, where you're weak, what your learning shape is. That memory drives tomorrow's plan.
What this category is good at: the long arc of a term. The "what should I do right now" question. Adapting when a midterm gets moved or you're sick for a week. Surfacing things you'd forget on your own.
What this category is bad at: being a fast lookup. If you just want a quick definition, you'll use ChatGPT. If you just want a problem solved, you'll use Wolfram. An integrated system is more setup than that, and it should be — because it's doing more.
The honest comparison: a chatbot is faster for a single question. An integrated system is dramatically better for a whole semester. They're not actually competing for the same job.
Fennie specifically
Since this is Fennie's blog, I won't pretend to be neutral. I'll just say what we do and don't do.
We do: courses, calendar, daily plan, chat, notes, quizzes, flashcards, essays, memory across all of it. The chat doesn't hand out answers — it asks what you tried, walks you through. The quizzes pull from your notes, not from a generic pool. The calendar reshapes the plan as deadlines move.
We don't: solve problems for you faster than Wolfram (we use chat that teaches, on purpose), and we don't replace Grammarly for line-level grammar checking. We're not trying to be the calculator. We're trying to be the system.
Free tier: chat with basic models, basic generation. Premium is 49.99/year and unlocks the better models, unlimited generation, and full memory and analytics. We don't think every student needs premium — some real students get by on free for an entire semester. We think it's worth the upgrade somewhere around the third week of any serious course load, when you start hitting generation limits or wanting the better models for the harder problems.
Three real archetypes
Generic comparison posts list features. The features are mostly the same. What's actually different is fit. Here are three students I've watched and what works for them.
The engineering undergrad
Sophomore. Calc III, Linear Algebra, Mechanics, a CS course, a lit elective. Heavy problem-set load.
What works: An integrated system as the spine, because the calendar and the daily plan are doing real work. Wolfram Alpha as a verification tool — paste a system of equations in, check the answer at the end of a derivation. ChatGPT or Claude in the rare moments where the chat tutor in Fennie isn't hitting on a concept and they want a different angle.
What doesn't work: Trying to use ChatGPT as the spine. The student ends up doing problem sets the night before, fast, with no retention, and gets crushed on exams.
The 1L
First semester of law school. Five courses, all reading-heavy, all moving fast.
What works: An integrated study system that handles the case-reading workflow — notes per case, chat for socratic-style review, flashcards on the rules and tests. Calendar awareness for the murder pace of cold calls and the eventual final.
What doesn't work: Generic chatbots. Law students have figured out, faster than most, that the value of law school is being able to think on your feet. A chatbot that hands you the answer to a hypothetical hurts you, and the cold call is going to expose it. The integrated system, with chat that asks what you'd argue first, is actively training the skill being tested.
The MCAT prepper
Junior, six months out from the test, balancing classwork with content review and full-length practice exams.
What works: Spaced repetition is non-negotiable. Calendar awareness for the test date is enormous — every plan from now until April should know that April is coming. Chat for working through high-yield concepts. Notes that turn into quizzes that turn into flashcards is the loop that gets MCAT content into long-term memory.
What also works: Anki, in the legacy Anki ecosystem. We're not pretending we replace Anki for the hardcore prepper who has tuned their AnKing deck. But for anyone starting fresh, an integrated system is more humane than maintaining decks by hand.
What doesn't work: Photomath. There are no Photomath problems on the MCAT. Some students try to use it for the physics section and just feel worse.
A simple decision tree
If you have one specific question right now and want an explanation, use a generic chatbot.
If you have a math problem you want solved or verified, use Wolfram or similar.
If you're writing and want grammar feedback, use a writing tool — but don't expect it to make the writing good.
If you have a semester to get through, with multiple courses, deadlines, and a real workload — use an integrated study system. The other tools will fit inside it as auxiliaries.
That's the actual answer. It's not "Tool X is best." It's "match the tool to the problem."
What I'd skip
Three categories I don't think students should spend time on:
Generic AI study apps that are really just ChatGPT wrappers. If the product's main feature is "AI chat for homework," you can do that with the free version of the underlying model. You're paying for nothing.
"AI flashcard generators" that don't sit inside a study system. Cards in a vacuum, with no spaced-repetition engine and no connection to what you're actually learning, are barely better than not making them.
Tools that promise to "do" homework — write essays, solve problem sets — and pretend that's a study workflow. That's not studying. That's outsourcing. The bill comes due at the exam.
The honest summary
Generic chatbots are great calculators of language. Specialized math tools are great calculators of math. Writing tools are great copy editors. None of these are study systems.
An integrated study system is a different shape — and for the actual job of getting through a hard semester, it's the shape that fits. Not because integrated systems are inherently better than chatbots. Because the job is bigger than what a chatbot can do.
If you're reading this in week one of the term, pick a primary system and let the others be auxiliaries. If you're reading this in week eight and feeling behind, the answer is the same — except start now instead of three months ago.