Most "best AI tools" lists are the same twelve products with affiliate links. This one is organized by the problem you're actually stuck on, because that's how you pick a tool.
If you're stuck on the math itself
Wolfram Alpha remains the most reliable computational engine available to students. It's not a language model, it computes symbolically, which means it doesn't hallucinate an answer the way a chatbot can. For checking whether an integral is right, it's still the standard.
Symbolab and Photomath show worked steps. The honest caveat: step-by-step solutions are excellent for checking your work and terrible for learning if you go there first. The difference between a student who improves and one who doesn't is usually whether they attempted the problem before looking.
Desmos and GeoGebra are free graphing tools that are genuinely better than most paid alternatives. If you can't picture what a function does, graph it.
If you're stuck on understanding a concept
This is where language models earn their place. ChatGPT and Claude are both strong at "explain this like I've already taken calc 1 but not calc 2", the specific kind of pitched explanation textbooks are bad at, because a textbook has to pick one level and stay there.
The technique that works: don't ask for an explanation. Ask the model to quiz you, then explain what you got wrong. Passive reading feels productive and mostly isn't.
The caveat that matters in STEM: language models are confidently wrong about arithmetic and edge cases. Use them for why, and use Wolfram for what.
If you're stuck on retention
Anki is unglamorous, free, and has more evidence behind it than anything else on this list. Spaced repetition works. The reason people abandon it is that making good cards is tedious, which is the one place AI genuinely helps. Generate draft cards, then edit them, because auto-generated cards you haven't reviewed are cards you won't remember.
If you're stuck on writing it up
Overleaf for LaTeX, with AI autocomplete that's now good enough to be worth the learning curve. For lab reports and problem sets, the bottleneck is usually formatting equations, not prose.
If your problem is access, not capability
This is the category most lists skip, and it's the one that's changed most.
If you're in a lecture or a lab and you have a question, your options are interrupt, write it down and forget it, or pull out a phone, which in most classrooms isn't allowed. The information exists. You just can't get to it in the moment you need it.
Dedicated devices exist for this. Nexus Pen is one: an AI pen with a built-in display where you press a button, ask out loud, and read the answer on the pen. It's a real ink pen, so it doesn't attract the attention a phone does. Its school mode is tuned to explain reasoning rather than just return an answer.
It's not a replacement for any tool above. Long derivations, file analysis, and extended back-and-forth all belong on a laptop. It's for the questions you have while your hands are already writing.
What to skip
Anything promising to do your homework. Beyond the academic integrity problem, the mechanism doesn't work, you cannot outsource the practice that builds the skill and still have the skill.
AI "study planners." A calendar works. These add friction and a subscription.
Anything that won't tell you when it's uncertain. In STEM, a tool that's confidently wrong is worse than one that's usefully unsure.
The short version
Wolfram for computation. A language model for concepts. Anki for retention. Desmos for intuition. Overleaf for writing. And something that works where your phone can't, if that constraint is real for you.
Five tools you actually use beat twenty you bookmarked.
Nexus Pen is an AI-powered pen with a built-in color display.