Getting an answer and learning something feel identical in the moment and diverge within a week. Several of these prompts are deliberately slower than just asking.
The most common thing people do with AI is ask it to help them understand something
How People Use ChatGPT — the NBER working paper by Chatterji, Cunningham, Deming, Hitzig, Ong, Shan and Wadman, built on a privacy-preserving analysis of 1.5 million conversations — found that three-quarters of all conversations fall into three categories: practical guidance, seeking information, and writing. Practical guidance, which includes tutoring and explanation, is the largest single category of use.
Most of it is not work. A substantial majority of usage is personal, and a large share of that is someone trying to understand something for themselves — a concept from a course, a topic they never learned, a language, a problem set they are stuck on.
This pack is built for that, and it is designed around one distinction that determines whether any of it helps: the difference between getting an answer and learning something.
Those feel identical in the moment and diverge completely a week later. A model that hands you a worked solution produces the sensation of understanding without the retrieval practice that creates it. So several prompts here are deliberately built to be slower than just asking — to make you attempt, retrieve, and self-test rather than read a clean explanation and move on.
The strictest accuracy rules on the site
Learning content is where a confidently wrong answer does the most damage, because you will carry it into an exam or a conversation and have no reason to doubt it. Every prompt in this pack carries three constraints:
Do not fabricate facts, dates, citations, or sources. A plausible-looking reference to a paper that does not exist is a real failure mode, and it is worse in study material than almost anywhere else.
Flag uncertainty rather than smoothing over it. Where something is contested, approximate, or outside what the model can reliably answer, it says so.
Self-verify before presenting. The quiz, exam, and problem-walkthrough prompts check their own answers before showing them, because a practice quiz with a wrong answer key teaches you the wrong thing and does it with the authority of an answer key.
Treat all of it as a well-read study partner rather than a textbook. For anything that will be assessed, check against your actual course material.
Understanding a thing
Explain a Concept at the Right Level is the entry point, and its value is the calibration: the same concept explained for someone with no background, for someone with adjacent knowledge, and for someone who needs the rigorous version are three different explanations, and getting the wrong one is why people conclude they are bad at a subject.
Explain It Like I'm Five is the deliberate floor — analogy-first, jargon-free, for when you need a foothold before anything technical can land.
Break a Complex Topic Into a Concept Map surfaces structure rather than content. Most difficulty with a large topic is not that any individual piece is hard; it is not knowing how the pieces relate, which makes everything feel like disconnected memorisation.
Learning it, rather than reading it
Study With a Socratic Tutor is the prompt that most embodies this pack's argument. It is instructed to guide rather than tell — to ask what you think, to respond to your reasoning, and to withhold the answer while you are still capable of reaching it. It is more effort than asking directly. That is the mechanism, not a side effect.
Walk Me Through a Problem Step by Step does the same for problem-solving, and finishes by generating practice problems of the same type — because following someone else's solution is not the skill you need on an exam.
Turn Notes Into Flashcards converts material into retrieval practice, which is among the best-evidenced study techniques there is and among the least used, because it is harder than re-reading and feels less productive while you do it.
Generate a Practice Quiz With an Answer Key and Take a Timed Practice Exam apply the same principle at scale. The timed version matters for exams specifically: knowing the material and performing under time constraint are separate skills, and the second one is only trainable by practising it.
Organising the work
Build a Study Plan for an Exam works backwards from the date and — importantly — from your honest available hours, so the plan is one you might actually follow rather than an aspirational schedule you abandon in week two.
Turn Lecture Notes Into a Study Guide restructures your own material rather than replacing it with generic content, which keeps it aligned to what your course will actually test.
Build a Self-Study Roadmap for a Skill is for learning outside any institution — a sequence with checkpoints, so you can tell whether you are progressing rather than only whether you are busy.
Writing and feedback
Outline an Essay or Paper is deliberately an outline prompt and not a drafting one. It develops your argument, structures it, and marks where you need evidence with a placeholder rather than supplying invented sources or finished prose. Handing in generated text is academic misconduct at most institutions, and this pack is written not to help with it.
Get Feedback on Your Work Against a Rubric works from the actual rubric you were given, which is far more useful than general writing advice.
Practice a Language Through Conversation is the pack's clearest win — unlimited low-stakes conversation at your level, with correction, which is precisely the thing language learners cannot otherwise get cheaply.
Teaching rather than learning? The teaching pack is the educator-facing counterpart — lesson plans, rubrics, assessments and differentiation.
Where this stops
These prompts do not know your syllabus, and they can be wrong. Nothing here guarantees a grade, a pass, or a timeline, and no prompt should be used to produce work you submit as your own. Check anything factual against your course materials, and check your institution's policy on AI use before you rely on these for assessed work — that policy varies enormously and the consequences of guessing wrong are serious.
Sources
- Aaron Chatterji, Tom Cunningham, David Deming, Zoë Hitzig, Christopher Ong, Carl Shan and Kevin Wadman, How People Use ChatGPT, NBER Working Paper 34255, September 2025 — 1.5 million conversations; practical guidance, seeking information and writing account for about three-quarters of use, with the majority of usage being non-work