Clear, practical technology insights BSOD Code Lookup · Windows Error Code Lookup · Wi-Fi Troubleshooting · PC Troubleshooting Checklist

ChatGPT Prompt Ideas for Writing, Coding, Study and Planning

Adapt practical prompts for drafts, debugging, schedules, research and decisions, with instructions that keep assumptions and unsupported claims visible.

Table of Contents

A useful ChatGPT prompt gives a clear task, relevant context, important constraints and the form of answer you need. You do not need a complicated role or a promise of “perfect” results. Give the model enough information to produce a draft you can evaluate, then refine the parts that miss your goal.

These examples adapt the writing, coding, planning and learning tasks often promoted for GPT-5. They are general prompt patterns, not a claim that GPT-5 is the latest available model or that every ChatGPT plan has the same tools. Replace the bracketed material with your own information, omitting private data you are not authorized to share.

Build a brief before choosing a prompt

Include the following when they matter:

  • Task: What should be produced or changed?
  • Context: Who will use the result, and what source material should guide it?
  • Constraints: What must stay, what must be avoided and what limits apply?
  • Output: Do you need a draft, table, code patch, explanation or list of questions?
  • Checks: Which claims, calculations or behaviors must be verified?

For example, “make this better” gives little direction. “Edit this customer notice to 150 words, preserve the refund conditions and make the next action clear” provides a result you can check. See writing context and a clear goal for a reusable approach.

1. Draft and edit writing

Outline a focused article

I am writing for first-time travelers who want to reduce waste without buying specialist products. Propose an outline for an 800-word article. Give each section a reader question and one practical example. Mark factual claims that need sources; do not invent statistics.

Review the outline for coverage and repetition before asking for a full draft. The audience and scope are more useful than a vague request for an “ultimate guide.”

Edit without changing the facts

Edit the text below for a clear, conversational tone. Preserve names, dates, qualifications and the original meaning. Do not add claims. After the revision, list any ambiguous sentence you could not safely resolve. Text: [paste draft]

Compare the revision with the source. A smoother sentence can still change the strength of a claim or accidentally remove an exception.

2. Generate and debug code

Specify behavior and boundary cases

Write a Python function named calculate_tax(income: float) -> float for this fictional exercise: apply 10% to the first 50,000 units of income and 20% only to income above that threshold. Reject negative inputs. Include a docstring and tests for 0, 50,000, 60,000 and a negative value. This is not a real tax rule.

Expected behavior matters more than asking the model to “act as a senior developer.” Here, a test at 60,000 should distinguish marginal calculation from incorrectly applying 20% to the entire amount.

Diagnose a reproducible error

Help diagnose this JavaScript failure. Runtime and version: [details]. Expected behavior: [description]. Actual behavior and full error: [details]. Minimal code: [code]. Identify the likely cause, propose the smallest fix and show how to verify it. If context is missing, say what would distinguish the possible causes.

Remove secrets before sharing logs. Run the proposed change in an appropriate test environment; an explanation that sounds plausible is not evidence that the fix works.

Understand unfamiliar code

Explain this C++ function for someone familiar with loops but new to pointers. Describe its inputs, output and side effects, then trace one small example. Flag behavior that depends on code not included here. Code: [paste code]

3. Plan a day or a longer project

Create a realistic schedule

Plan tomorrow from 9 am to 5 pm using these tasks and duration estimates: [list]. Fixed appointments: [list]. Include lunch, short breaks and a buffer. I prefer demanding work before lunch. If everything will not fit, identify what to defer rather than compressing the estimates.

Check dependencies and commitments yourself. A generated schedule is a proposal; it does not mean calendar events or reminders have been created.

Break down a large goal

I have 12 weeks to complete a 50-page thesis draft. I can work [hours] weekly. Current stage: [details]. Build milestones for research, outline, drafting and revision. Include supervisor feedback time and identify assumptions that could make the plan unrealistic.

Adjust the plan to your institution's requirements. Page count is a workload cue, not a substitute for the actual research deliverable.

4. Draft a practical email

Draft a polite, firm email to my internet provider about these documented outages: [dates and durations]. Ask for an explanation and a review of any credit available under my service terms. Use placeholders for my account number and contact details. Do not invent an entitlement or send the email.

Supply the facts you can support and fill in sensitive account information outside the chat if appropriate. Before sending, verify dates, the requested action and any statement about the service agreement.

5. Study through explanation and practice

Ask for an explanation with limits

Explain thermodynamic entropy to a beginner using one everyday analogy. Then explain where that analogy breaks down. Include the meaning of the term in the scientific context and one short example that checks my understanding.

Asking where an analogy fails can prevent a memorable comparison from becoming a misleading definition.

Practice actively

Using only the class notes below, ask me five questions about the French Revolution, one at a time. Wait for each answer, explain the relevant evidence from the notes and revisit concepts I miss. Do not introduce unsupported dates or quotations. Notes: [paste notes]

For a Socratic discussion, ask for one question at a time and a clear learning objective. Verify important explanations against your course material, especially when the model departs from the provided notes.

6. Develop business ideas without invented forecasts

Compare small experiments

I run an online bakery serving [area and customer type]. Constraints: [budget, production capacity and delivery limits]. Suggest five low-cost experiments to test demand next quarter. For each, give a hypothesis, cost category, success measure and reason to stop. Do not predict a guaranteed sales increase.

Draft marketing copy from approved facts

Write a 100-word Instagram caption for this product using only these approved details: [details]. Audience: [description]. Tone: friendly and specific. Include one call to action. Do not invent health benefits, customer testimonials, certifications or scarcity.

For market or SWOT analysis, provide current evidence and ask the model to distinguish sourced observations from assumptions. A plausible market narrative is not market research. The business prompting guide gives additional department-specific examples.

7. Compare decisions and stress-test plans

Create a decision matrix

Compare Product A and Product B using the data below. Criteria: development cost, time to launch, demand evidence and downside risk. Keep unknown values marked Unknown. Show how the recommendation changes under different priorities instead of inventing missing scores. Data: [details]

Explore a supplier-price shock

Suppose our main supplier doubles its price. Using this cost breakdown, compare finding another supplier, raising prices and reducing other costs. Show assumptions, tradeoffs and information needed before acting. Do not present the scenario as a forecast. Cost breakdown: [details]

Use the comparison to identify what you still need to learn. Check arithmetic and feasibility before committing money or making operational changes.

8. Explore philosophy and creative ideas

Compare philosophical positions

Compare moral absolutism and moral relativism on this dilemma: [scenario]. Explain each position charitably, identify a strong objection to each and distinguish the theories from particular philosophers' views. Do not invent quotations or citations.

If you simulate a debate, make the assigned positions explicit. An AI's argument for a view is not proof that the view is true or that the system personally holds it.

Build a story under constraints

Write a 600-word detective story about a cat investigating missing milk. Use a restrained noir tone, three plausible clues and a solution supported by those clues. Avoid explaining the moral at the end.

Create three distinct concepts for a music-powered superhero living in a floating city. For each, describe a limitation, a personal conflict and a setting detail that affects the plot. Avoid copying an existing character.

Creative constraints give you alternatives to judge. Pick a promising direction and request specific revisions instead of repeatedly asking for “more creativity.”

9. Summarize and extract from sources

Summarize a paper without losing its limits

Summarize the attached paper's research question, method, main findings and limitations. Distinguish measured results from the authors' interpretation. Give page or section references where available. If the attachment is incomplete or unreadable, state that rather than filling gaps.

Do not ask the model to generate missing research data and then treat it as evidence. Label synthetic examples clearly if you use them for practice.

Extract interview evidence

From this transcript, extract passages about remote-work productivity into a table with theme, exact short quotation and speaker or timestamp. Separate direct quotes from paraphrases. If fewer than five relevant passages exist, return only those present. Transcript: [paste text]

Check quotations against the original recording or transcript. For research design, ask for questions and methodological tradeoffs to discuss with a supervisor; a chatbot proposal does not replace ethics approval or expert review of a study involving people.

Refine the result with specific feedback

Use a follow-up such as “keep sections 1 and 3, remove the repeated example in section 2 and mark every unsupported factual claim.” For complex work, approve an outline or data structure before expanding the draft, then verify the finished output against the original requirements.

OpenAI's archived GPT-5 prompting guide emphasizes clear, consistent instructions and experimentation; its model and API details belong to that guide's period. For everyday use, measure success by the usefulness and correctness of the result. The writing-quality checklist can help turn a generated draft into work you can rely on.

Discussion

Reader Comments 0

Sign in with email or Google to join the discussion.