Every prompt below is a template you can lift straight into ChatGPT, Claude, Gemini or any other model. The parts in [square brackets] are yours to fill in: swap them for your real topic, audience, product or tone, delete any line that doesn't apply, and add one sentence of your own context so the model isn't guessing. If a first answer misses, don't rewrite from scratch. Reply with what to change ("shorter", "more concrete examples", "drop the intro") and let the model revise. For the reasoning behind these patterns, see how to write AI prompts.
Writing & content
These give the model a role, a format and a length so you get a usable draft instead of a wall of text.
Rewrite the text below so a [busy manager / 12-year-old / non-native English reader] can understand it in under 30 seconds. Keep every fact. Use short sentences and no jargon. Text: [paste text]Write a [200]-word [blog intro / product description / cover letter] about [topic]. Audience: [who]. Tone: [confident but plain]. Open with a specific detail, not a definition. End with one clear call to action.Turn these rough notes into a structured [email / summary / report]. Group related points under headings, cut repetition, and flag anything that looks contradictory. Notes: [paste notes]Give me [5] alternative headlines for [article/topic]. Vary the angle: one question, one how-to, one with a number, one contrarian, one plain. No clickbait, no exclamation marks.Act as a line editor. Tighten the paragraph below without changing my voice or meaning. Show the edited version first, then a short bullet list of what you changed and why. Paragraph: [paste]Draft a [reply / follow-up] to this message. Goal: [what I want to happen]. Tone: [polite and firm]. Keep it under [120] words. Original message: [paste]
Marketing & social media
Name the platform and the goal. "Write a post" is vague; "a LinkedIn post that gets comments" gives the model something to optimize for.
Write [3] variations of a [LinkedIn / Instagram caption / X] post announcing [product or news]. Audience: [who]. Hook in the first line. One idea per post. Include a soft call to action and [2-3] relevant hashtags where they fit.I sell [product] to [customer]. List [10] content ideas built around their real problems, not our features. For each: the pain point, the format (how-to, checklist, story, comparison), and a one-line hook.Write a [4]-email welcome sequence for someone who just signed up for [product/newsletter]. For each email give: goal, subject line, and 3-4 sentences of body. Sell nothing until email [3].Rewrite this ad copy for [Google Search / Meta]. Fit [30] characters for the headline and [90] for the description. Lead with the benefit, include [keyword], and make the call to action specific. Current copy: [paste]Give me [5] short-form video hooks (first 3 seconds) for a clip about [topic]. Each should create curiosity or state a bold-but-true claim. Add a one-line idea for the payoff.Draft a customer-facing announcement for [change/feature]. Lead with what's in it for them, keep the "why" to one sentence, and reassure anyone worried about [likely concern]. Under [150] words.
Coding & development
Say the language, the versions and what "done" looks like. Ask for an explanation and edge cases, not just code you'll paste without reading.
Write a [Python] function that [does X]. Inputs: [types]. Output: [type]. Handle [empty input / bad values / timeouts]. Add type hints and a short docstring. Then show [2] example calls with their output.Here is an error and the code that caused it. Explain the likely cause in plain terms, then give the smallest fix. Don't refactor anything unrelated. Error: [paste]. Code: [paste]Review this function for bugs, edge cases and readability. List issues by severity (blocker, should-fix, nice-to-have). Show a corrected version only after the list. Code: [paste]Explain what this code does line by line, as if onboarding a new teammate. Point out anything surprising or risky. Assume they know [language] but not this codebase. Code: [paste]Write [framework] tests for the function below. Cover the happy path, boundaries, and [2] failure cases. Use [pytest / Jest] and name each test after the behavior it checks. Function: [paste]Convert this [SQL query / snippet] from [source] to [target]. Keep the behavior identical, note any differences that could change results, and don't add features I didn't ask for. Code: [paste]
Study & learning
Ask the model to teach, quiz and correct you. Passive summaries are forgotten fast; retrieval and feedback are what stick.
Explain [concept] at three levels: one sentence for a beginner, a paragraph for a student, and the precise version for an expert. End with the single most common misconception about it.I have [3] days to learn [topic] for [exam/purpose]. Build a day-by-day plan with what to study, one practice task per day, and how I'll know I've got it. Assume [1-2] hours a day.Quiz me on [topic] with [8] questions, one at a time. Wait for my answer before the next. After each, tell me if I'm right, correct me briefly, and get slightly harder if I'm doing well.I'll explain [concept] in my own words; you find the gaps. Point out anything wrong, vague or missing, then give me one question that tests whether I really understand it. My explanation: [paste]Turn this material into [10] flashcards as a two-column table (question, answer). Keep answers to one line. Focus on things I'd be tested on, not trivia. Material: [paste]Give me [2] worked examples for [type of problem], solved step by step with the reasoning shown, then [2] more for me to try with the answers hidden until I ask.
Business & productivity
Feed the model your real constraints (budget, deadline, audience) and ask for a decision or a draft you can act on, not generic advice.
Summarize the notes below into [decisions made], [action items with an owner], and [open questions]. If an owner or a due date is missing, mark it TBD. Notes: [paste]Turn this goal into a plan: [goal] by [date]. Give me the [3-5] milestones in order, the first task under each, and the one thing most likely to derail it.Draft an agenda for a [30]-minute meeting about [topic] with [who]. Time-box each item, put the decision that must be made first, and end with clear next steps.Act as a skeptical reviewer of this plan. List the [5] biggest risks or weak assumptions, and for each suggest one cheap way to test it before we commit. Plan: [paste]Compare [option A] and [option B] for [decision] as a table: criteria in rows (cost, time, risk, effort), options in columns. End with a one-line recommendation and the main trade-off.Write a clear, friendly reply declining [request] without burning the relationship. Give a brief honest reason, offer one alternative, and keep it under [80] words. Request: [paste]
Image prompts (Midjourney / DALL·E)
Image models reward description over instruction. Stack subject, setting, style, lighting and framing, then adjust one thing at a time. There's a fuller walkthrough in Midjourney prompts.
[subject], [doing what], in [setting], [time of day and weather], [art style or medium], [lighting], [camera or framing], [mood]. Example: A ceramic coffee cup on a windowsill, morning light, soft shadows, minimalist still-life photography, shallow depth of field, warm and calm.Product shot of [product] on [surface/background], studio lighting, [angle: 3/4 view / top-down], crisp focus, subtle reflection, clean negative space for text, commercial photography.Flat vector illustration of [subject], [2-3] color palette ([colors]), simple geometric shapes, bold outlines, white background, suitable for an app icon.Character concept: [who], [age and build], wearing [outfit], [expression], [pose], [environment behind them], semi-realistic digital painting, cinematic lighting, full-body.Wide landscape of [place], [season], [atmosphere: misty / golden hour], painterly style inspired by [genre, not a living artist], high detail, dramatic sky.[scene], isometric view, cozy diorama, soft studio lighting, muted pastel palette, tilt-shift look, high detail. Add at the end for Midjourney: --ar 16:9 (or --ar 1:1 for a square).
Whichever category you start from, the habit is the same: give the model a role and a format, fill the brackets with your real details, and refine by reacting to the draft. When you have the shape of a prompt you like, drop it into the prompt generator to tidy the wording and adapt it to the exact model you're using.