AI Prompt Examples: 30 Templates That Actually Work
A good example beats a long explanation. Thirty ready-to-copy AI prompt examples across work, study and creativity — and the pattern behind each one.
You don't need to be clever to get good output from an AI. You need a good starting point. The fastest way to learn what a strong request looks like is to copy one that works, run it, then bend it to your task. That is what this page gives you: a shelf of AI prompt examples you can lift straight into ChatGPT, Claude or Gemini, grouped by the job you're trying to do. Each comes with a one-line note on why it works.
The anatomy every good example shares
Skim the prompts below and you'll notice they share four moving parts. Learn these and every example is easy to adapt.
- Role — who the AI should act as ("you are a copy editor"). Pulls in the right vocabulary.
- Task — the actual verb: write, summarise, compare, refactor, plan. One clear ask beats three vague.
- Context — the background only you know: audience, constraints, source text.
- Format — what the answer should look like: five bullets, a table, valid JSON, a 200-word email.
The best AI prompt examples are those four boxes filled in well. When a result is wrong, one box was empty. The guide on how to write AI prompts covers each.
Writing and content
The length target does the work; "make it tighter" doesn't.
"Concrete scene" kills the throat-clearing opening AI defaults to.
Asking for distinct angles gets real variety, not ten rewordings.
Scoping to three lines keeps feedback usable, not a wall of notes.
A different idea per post stops near-identical summaries. More in AI prompts for writing.
Work and productivity
Forbidding invention stops the model padding with filler.
The fixed action-item shape drops into a tracker.
The tone pair prevents the over-apologetic default.
Naming the words to cut beats "make it more assertive".
Asking first builds the plan on your real priorities, not a guess.
Save the ones you reuse in a prompt library so you're not rebuilding the brief each Monday.
Coding
Making it state assumptions surfaces bugs before you paste anywhere.
The severity ranking tells you what to fix first.
The "at scale" clause catches landmines in inherited code.
Ranking by likelihood beats a list of everything possible.
The per-part comments let you trust a regex you didn't write. More in AI prompts for coding.
Marketing
Banning the default promo words removes most of the mush.
One job per email stops it repeating the value prop.
Arguing with the outline first beats one long generation.
Repurposing by claim gives posts that don't sound identical.
Written after the body, the lines pull in real details. The marketing prompts guide shows how to add your own proof.
Study and learning
The ladder shows where your understanding drops off.
"One at a time, wait" turns a list into a real session.
The daily self-test makes it stick.
The "commonly get wrong" section pre-empts your mistakes.
Teaching it back and asking for holes finds them fastest.
Image generation
Image prompts stack subject, style, lighting and framing, and name the tool. Keep them in English — the models learned mostly English tags.
For Midjourney — the --ar flag and film cue steer it cinematic.
For DALL·E — "flat vector" and "minimal background" stay web-hero clean.
For Stable Diffusion — lens and film stock push toward believable skin.
Works across tools — "isometric cutaway" gives a tidy scene.
For Stable Diffusion or DALL·E — "seamless background" and "diffused lighting" fake a photo booth.
The image prompt builder assembles those tags for you, and the Midjourney prompt generator handles the flags.
How to adapt an example to your work
Every prompt above is a template, not a script. To adapt one, work the four boxes in order:
- Swap the context. Replace the placeholder product, audience or source with your real one — most of the quality gap lives here.
- Tighten the format. Change "under 120 words" or "5 bullets" to what you need. Numbers give you control.
- Adjust the role. If the tone is off, the role is usually why — "friendly coach" and "blunt reviewer" are worlds apart.
- Add a constraint the model can't guess. A banned word, a fact it must keep, a thing to avoid.
The strongest AI prompt examples are the ones you've edited twice — each edit shows what the model was guessing at. If you would rather not build them by hand, feed a plain request to the prompt optimizer and it adds the role, constraints and format for you, or start from the AI prompt generator. Ready-made prompt templates tuned for one model start with the ChatGPT prompt generator.
Copy a few, break a few, and soon you won't need examples at all.
