What Is a Prompt? A Plain-English Guide with Examples
A prompt is simply the instruction you give an AI. Get the instruction right and the answer
follows — here’s what a prompt is, what makes one good, and a few you can copy and try.
A prompt is the text you type to tell an AI model — ChatGPT, Gemini, Claude
— what you want it to do. That’s the whole idea. It can be as short as a question (“What’s a
good gift for a 7-year-old?”) or as detailed as a full brief that sets a role, context, a
task, the format you want back, and rules to follow. The model reads your prompt and predicts
the most useful response it can. So the quality of what you get out is mostly decided by what
you put in.
That’s why “write me a blog post” returns bland filler while a specific, well-structured
prompt returns something you can actually use. You don’t need to be technical to write a good
one — you just need to include a few parts most people leave out.
The anatomy of a good prompt
Strong prompts tend to include five things. You won’t need all five every time, but naming them is what turns a vague request into a reliable one:
Role — who the AI should act as. “You are an experienced copy editor…”
Context — the background and any material to work from. “Here are my rough notes: …”
Task — the one thing you want done. “Rewrite this for clarity.”
Format — how the answer should look. “Reply as five bullet points, under 100 words.”
Guardrail — what not to do. “Don’t change my meaning or invent statistics.”
A weak prompt vs. a strong one
Same goal, very different results:
Weak
“Write a cold email to a sales prospect.”
No role, no context, no constraints — you get a generic template anyone could have written.
Strong
“You’re an SDR. Write a 90-word cold email to a marketing director at a 50-person SaaS company who just posted about hiring their first content lead. Open with that observation, make one specific point, and end with a single low-friction question. No buzzwords.”
Role, context, task, length, format and a guardrail — the output reads like a real message.
Prompt examples to copy
Here’s what a complete prompt looks like in practice. Each one has fill-in [BRACKETS] you edit before copying, plus an example of what it returns:
ChatGPTGeminiClaude
Turn messy meeting notes into clear action items
Convert raw notes or a transcript into decisions, owners, and next steps.
The prompt 1 field to fill
Below are raw notes from a team meeting. Turn them into a clean summary for people who weren't there.
Format:
1. **TL;DR** — 2 sentences.
2. **Decisions made** — bullet list.
3. **Action items** — a table: Task | Owner | Due date. If an owner or date wasn't stated, write "unassigned" — do not invent one.
4. **Open questions** — anything left unresolved.
Keep it factual. Don't add advice. Here are the notes:
Example output
**TL;DR** — The team approved the Q3 pricing test and pushed the mobile redesign to August. Two action items are unassigned.
**Decisions made**
- Run the pricing A/B test on the Pro plan starting July 1.
- Delay the mobile redesign to August to avoid the holiday code freeze.
**Action items**
| Task | Owner | Due date |
|---|---|---|
| Set up pricing test in billing | Priya | Jun 27 |
| Draft August launch timeline | unassigned | unassigned |
| Tell support about the price change | Marco | Jun 30 |
**Open questions**
- Who signs off on the final August date?
TipThe "do not invent owners or dates" rule is what makes this trustworthy — AI otherwise fills gaps with plausible-but-wrong names. Paste a Zoom/Meet transcript directly for best results.
ChatGPTGeminiClaude
Summarize a research paper
Get the key claims, method, and findings of a dense paper.
The prompt
Act as a research methods tutor. I will give you the abstract and any sections I paste from a research paper. Summarize it using this structure:
1. Research question (one sentence)
2. Method (what they did, how many participants or data points)
3. Key findings (bullets, with numbers if they appear)
4. Limitations the authors mention
5. One sentence on why this paper matters
Do not interpret or add analysis. Use plain language I can explain to a classmate.
Paper excerpt: [PASTE ABSTRACT + KEY SECTIONS]
Example output
1. Research question: Does spaced repetition improve long-term retention compared to massed practice in undergraduate vocabulary learning?
2. Method: Randomized experiment, 94 undergraduate psychology students over 8 weeks; one group used a spaced flashcard app, one group did blocked study sessions.
3. Key findings:
- Spaced group retained 31% more words at the 4-week follow-up test.
- No significant difference at the immediate post-test (p = .42).
- Effect was strongest for low-frequency words.
4. Limitations: Self-selected study time; both groups used the same word list.
5. Why it matters: Confirms spacing benefit persists beyond the lab and applies to a practical student tool.
TipIf the paper is long, paste the abstract and the Results section first. Those two alone give you 80% of the structure; add the Discussion only if you need the "why it matters" layer.
MidjourneySDXLFLUX
Cinematic close-up portrait
A film-still style headshot with controllable subject, mood and lighting.
The prompt 4 fields to fill
cinematic close-up portrait of , expression, shot on 85mm f/1.4 lens, shallow depth of field with creamy bokeh, lighting, color grade, fine skin texture and catchlights in the eyes, photorealistic, 35mm film grain --ar 4:5 --style raw
Example output
A tight head-and-shoulders portrait of a weathered fisherman in his 60s, a faint knowing smile, lit by warm low-key window light from the left that falls off into shadow on the right cheek. The teal-and-amber grade and visible film grain read like a frame pulled from an indie drama; the eyes are sharp with twin catchlights while the background dissolves into soft bokeh.
TipSwap in any subject and keep the lens + lighting + grade scaffolding — that trio is what makes it read as a cinematic still rather than a generic AI face. On Midjourney, `--style raw` reduces the over-polished default look; in SDXL/FLUX drop the `--ar/--style` flags and set aspect ratio in the UI.
A prompt is the instruction you give an AI model like ChatGPT, Gemini or Claude to tell it what you want. It can be a single question (“Explain photosynthesis simply”) or a detailed brief with a role, context, the task, the output format, and rules. The clearer and more specific the prompt, the more useful the response.
What is an example of a good prompt?
A good prompt names a role, gives context, states the task, specifies the format, and adds a guardrail — for example: “You are a hiring manager. Using the notes below, write a 120-word rejection email that’s warm and final. Don’t invent any feedback I didn’t give.” The copy-ready examples on this page all follow that pattern.
What’s the difference between a prompt and prompt engineering?
A prompt is the instruction itself. Prompt engineering is the practice of writing, testing, and refining prompts to get reliably good results — choosing the right structure, examples, and constraints. You don’t need to be a prompt engineer to write a good prompt; you just need the few habits covered above.
Do prompts work the same in ChatGPT, Gemini and Claude?
Mostly, yes — the same well-structured prompt works across all three. There are small differences: Gemini is handy when you need current sources, and Claude is strong with very long documents. The prompts in this library are written to work in all three unless a card notes otherwise.