ChatGPT Custom Instructions for University Students

Custom instructions let you set a persistent study mode for ChatGPT without repeating your preferences at the start of every session. For university students, the most useful instructions cover source verification, tutoring style, assignment boundaries, and subject-specific constraints.

The examples below are starting points you can adapt to your own institution's rules and your own courses. Tools like fenbs store your evolving set of reusable instructions and course notes so they carry across study sessions and AI apps.

What to tell ChatGPT about your role as a student

The first field in custom instructions is for background about yourself. Being specific about your level and field helps ChatGPT calibrate its explanations and avoid pitching concepts too high or too low.

I am a second-year undergraduate studying [your subject] at [your institution]. I want explanations that assume I know [topic I know well] but not [topic I am still learning]. I am studying in the United States and follow APA citation style.

Keep this field factual and brief. The second field is where you put behavioral rules; mixing the two makes both less effective.

Source checking and citation instructions

One of the most important rules for academic use is that ChatGPT should not invent sources. It can hallucinate plausible-sounding citations that do not exist. A clear instruction reduces this risk and prompts the model to flag uncertainty rather than fill gaps with invented references.

Never invent citations, authors, or study titles. If you are not certain a source exists, say so plainly and suggest I search for it myself. When I ask you to help with citations, ask me for the source details rather than generating them.

This instruction does not eliminate the hallucination risk entirely, but it prompts ChatGPT to hedge rather than fabricate. Verify any citation you plan to use in an assignment against your library catalog or a database like Google Scholar or JSTOR.

Tutoring mode instructions

If your goal is to understand a concept rather than to get an answer, an instruction that puts ChatGPT into a tutoring mode can make sessions much more productive. The Socratic approach — answering a question with a question — is well-supported by the model.

When I ask you to explain a concept or help me understand something, ask me what I already know before explaining. Use examples from [your subject area] where possible. If I give a wrong answer, explain why it is wrong before giving the correct one.

You can turn this off for specific conversations when you need a direct answer rather than a teaching dialogue. Adding "just answer directly" to a message overrides the rule for that exchange.

Assignment boundary instructions

Different institutions and instructors have different policies on AI use. FERPA governs how student records are handled in the US, and submitting AI-generated text as your own work is an academic integrity issue at most universities. An instruction that keeps ChatGPT in an advisory rather than a writing role helps you stay within those boundaries.

When helping with assignments, give feedback, outlines, and suggestions — do not write sections for me to submit. If I ask you to write something I will submit as my own work, remind me of academic integrity rules and offer to help me write it myself instead.

This is not a substitute for knowing your institution's specific policy. Check your student handbook or ask your instructor if you are unsure what is permitted for a particular assignment.

Subject-specific instructions

General instructions work for most sessions. For a subject with specific conventions, a targeted rule keeps the output aligned with what your discipline expects.

• Law: "Use US legal terminology. Cite cases in Bluebook format when I ask for citations. Flag when a legal principle varies by state."
• Medicine or health sciences: "Do not give medical advice. Frame clinical scenarios as learning exercises. Use the US spelling and classification system."
• Social sciences: "Use APA style by default. When discussing statistics, explain what they mean in plain terms as well."
• Engineering: "Show your working in calculations. Use SI units unless I specify otherwise."
• Writing and literature: "Do not summarize a text I have not given you. Ask for the passage before analyzing it."

Add subject-specific rules to your second field alongside your general behavioral rules. Keep the total under the character limit by being brief: one sentence per rule is usually enough.

Keeping your instructions and course references organized

Custom instructions in ChatGPT are one global set. If your rules change between courses, or if you want to keep a longer version of each rule alongside notes about why you wrote it, a separate tool makes that easier to manage.

How to do this with fenbs

1. Create a free account. fenbs is free for one person with no credit card required.
2. In Skills and prompts, create a skill for each course or study context. Paste your full instruction set, not just the compressed version that fits in ChatGPT.
3. Add a task for each assignment you are tracking, with a note on where you are and what is still to do.
4. Connect ChatGPT, Claude, or another supported AI app from Integrations. Ask it to read your study skills at the start of each session.
5. When a rule stops working or a new course needs different instructions, update the skill in fenbs and the change applies to every future session.

fenbs is not a plagiarism checker and does not enforce your institution's academic integrity rules. It is a place to organize your own instructions and project progress. Responsibility for following your university's AI use policy stays with you.

For the broader question of how to set up custom instructions in ChatGPT and what the two fields do, see ChatGPT custom instructions. If your instructions are saved but not being followed, ChatGPT custom instructions not saving covers the troubleshooting steps. For managing instructions across multiple AI apps, the ChatGPT Projects vs custom GPTs comparison explains when Projects make more sense than global custom instructions.