Best AI Prompts for Students

AI summary

Seven AI prompts that turn a study session from “AI did my work” into “I learned faster.” Each prompt is designed to scaffold thinking, not bypass it: lecture-to-study-sheet, assignment breakdown, Feynman self-test, source triage, outline critique, exam-prep pulse check, office-hours question builder.

Prompt 1

Lecture-to-Study-Sheet

After class, turn raw lecture notes (or an auto-generated transcript) into a study sheet you can actually review. The key move is asking for the structure you want, not just “summarize.”

I am attaching my notes from today's lecture on [TOPIC] in [COURSE].

Produce a study sheet with three sections:

1. CORE CONCEPTS: 5-7 bullet points, each one a precise definition the professor would accept on an exam.
2. WORKED EXAMPLE: One example problem from the lecture, fully solved with reasoning written out as if explaining to a classmate.
3. SELF-TEST: 8 questions I should be able to answer cold, in increasing difficulty. Hide the answers below a divider so I can quiz myself first.

Do not invent material that was not in my notes. If something is unclear in my notes, flag it so I can re-check with the professor.

When to use: Same-day or next-day review while the lecture is still warm. · Best model: Claude (best for following structural instructions) or ChatGPT.

Prompt 2

Assignment Breakdown

Most stuck-on-a-paper situations are actually “I have not broken this assignment into stages” situations. This prompt makes the AI do that work for you.

Here is the assignment prompt: [PASTE FULL ASSIGNMENT TEXT]

Due date: [DATE]. Today is [TODAY'S DATE]. I have roughly [X] hours per day available.

Do the following:

1. Identify the actual deliverable in plain English (not the professor's wording).
2. Break it into 5-8 stages, smallest first, with a realistic time estimate for each.
3. Suggest one specific milestone for each stage that I can check off so I know I made progress.
4. Flag the one stage where most students get stuck and explain how to spot if I am getting stuck there.

Do not write any of the assignment for me. I want the plan, not the work.

When to use: Day you get the assignment, before you open a blank doc. · Best model: Any frontier model. Claude is especially careful about not crossing the “do the work for me” line.

Prompt 3

Feynman Self-Test

Richard Feynman’s technique: you do not understand something until you can teach it to a 12-year-old. This prompt makes the AI grade your re-explanation.

I am going to explain [CONCEPT] from [COURSE] as if I were teaching it to a smart 12-year-old. Here is my explanation:

[YOUR EXPLANATION, written out fully in your own words]

Grade it on these criteria:

1. Accuracy: Did I get any technical detail wrong?
2. Completeness: What did I leave out that a 12-year-old would actually need to follow this?
3. Jargon creep: Where did I use a word I would have to define if pushed?
4. Mental hooks: Did I use any analogies or examples? If not, suggest one I could add.

Be specific. Quote my actual sentences when pointing out problems.

When to use: Night before the exam, on every concept you are not sure you really understand. · Best model: Claude or ChatGPT. Grok works for STEM topics where you want a more challenging critic.

Prompt 4

Research Source Triage

You have 15 sources for a paper. You will read 4. This prompt picks the 4 in priority order so you stop randomly opening PDFs.

I am writing a paper on [TOPIC] for [COURSE]. My thesis is: [ONE-SENTENCE THESIS].

Here are the sources I have collected (title, author, year, and a one-line description for each):

[PASTE LIST OF 10-15 SOURCES]

Rank these in priority order for my thesis. For each source in the top 4, tell me:

1. Why this source supports or complicates my thesis.
2. What specific question to look for when reading it.
3. How much time it deserves (skim, careful read, full deep read).

For the bottom sources, give a one-line reason why they are lower priority. Do not pretend a source is more relevant than it is.

When to use: After you have collected sources, before you start reading. · Best model: Claude (strong at hierarchical analysis) or any model with web search if you want it to verify source quality.

Prompt 5

Essay Outline Refiner

AI rewrites of your essay get you nowhere. Critiques of your outline get you somewhere. This prompt is the difference.

Here is my outline for a [LENGTH]-word essay on [TOPIC]:

[PASTE YOUR ROUGH OUTLINE]

Do not rewrite it. Instead:

1. Identify the strongest argument in the outline and explain why it is strong.
2. Identify the weakest section and explain what would have to be added to make it work.
3. Flag any section that is doing two jobs (covering two arguments) and suggest how to split it.
4. Identify one place where a counter-argument should appear that I have not included.
5. Suggest one specific source type (not a citation, a category) that would strengthen the weakest section.

Be direct. I would rather hear about the weak spots now than after the professor finds them.

When to use: Right after you have a rough outline, before you start drafting. · Best model: Claude is best for this. It is naturally less effusive than ChatGPT and will actually point out weaknesses.

Prompt 6

Exam Prep Pulse Check

You have a few hours and uneven knowledge. This prompt creates the targeted study plan instead of you re-reading what you already know.

I have an exam in [COURSE] on [DATE]. I have [HOURS] available between now and then.

Here is the syllabus topic list:

[PASTE TOPICS]

For each topic, here is my self-rating: solid / shaky / blank.

[Topic 1: solid]
[Topic 2: shaky]
[Topic 3: blank]
...

Produce a study plan that:

1. Spends 60 to 70 percent of my time on the shaky topics, not the blanks.
2. Allocates a single focused session for each blank topic with the smallest viable goal (recognize the concept, not master it).
3. Allocates a 30-minute review block for each solid topic at the END of the plan to catch decay.
4. Includes one full mixed-practice block in the final 24 hours.

Return it as a time-blocked schedule.

When to use: Two to three days before the exam, not the night before. · Best model: Any model. Claude for the most disciplined time-blocking.

Prompt 7

Office Hours Question Builder

Walking into office hours saying “I don’t get problem 3” wastes the professor’s time and yours. This prompt drafts a question precise enough to get a useful answer.

I am stuck on this problem from [COURSE]:

[PASTE PROBLEM]

Here is what I have tried:

[DESCRIBE YOUR ATTEMPTS, INCLUDING THE WRONG ANSWERS YOU GOT AND WHY YOU THINK THEY WERE WRONG]

Draft a question I could ask in office hours that:

1. Shows I have actually engaged with the problem (referencing my attempts).
2. Identifies the specific concept I think I am missing.
3. Asks a focused question that takes less than two minutes to answer.
4. Avoids asking the professor to solve the problem for me.

If my attempts reveal a deeper conceptual gap, point that out separately so I can flag it to the professor.

When to use: Night before office hours, after you have tried the problem yourself. · Best model: Claude is well-suited to this because it tends to push back rather than just solving for you.

Connectors are now standard

Claude, ChatGPT, and Grok all support connectors that let your AI read live data from your work tools (Gmail, Notion, GitHub, Asana, HubSpot, Stripe, and many more) instead of relying on you to paste context. For students this means the AI can read your Google Drive folder for that class, your Notion course notes, your Calendar to know what assignment is due next, or your Gmail to find the professor’s clarification email.

For students, the connectors worth pairing with these prompts:

Get Smarter About AI Every Morning

Free daily newsletter. Built for people who want to use AI well, not chase every model.

Free forever. Unsubscribe anytime.