How do you write better prompts? Stop hunting for magic phrases. Behind every good prompt sits a small set of repeatable patterns: give context, show an example, set constraints, specify the format, and iterate. Master these 15 and you can get reliable output from any AI assistant, on almost any task.

Each pattern below includes a copy-paste example and one line on when it shines. Adapt the bracketed parts to your situation.

1. Role plus task

Tell the assistant who to be, then what to do. A role tunes its vocabulary and standards to the task at hand.

You are an experienced hiring manager at a mid-size company.
Write a rejection email to a candidate who interviewed well
but lacked one required skill. Warm, honest, under 120 words.

Shines when: the task has professional standards you want the output to match.

2. Context dump

Paste the background material first, then ask. The assistant cannot read your mind, but it reads what you paste very well.

Here are my raw notes from yesterday's client call:

[paste notes]

Turn these into a follow-up email summarizing what we agreed, the three open questions, and who owns each one.

Shines when: the answer depends on information only you have.

3. Format specification

Describe exactly how the answer should look. Length, structure, tone. Vague format requests get vague-shaped answers.

Explain how our refund policy works in three short paragraphs.
Plain language, no legal jargon, friendly tone.
End with one sentence telling the reader what to do next.

Shines when: you know what good looks like and want it on the first try.

4. Few-shot examples

Show two or three examples of what you want, then ask for more in the same style. This is the most reliable pattern for matching a voice.

Here are two product descriptions in our brand voice:

1. [paste example]
2. [paste example]

Write a third one for [new product], same voice, same length.

Shines when: matching an existing style matters more than the content itself.

5. Constraints list

Give explicit do and do not rules. Constraints force decisions the assistant would otherwise dodge.

Draft a job post for a customer support lead.
Do: mention the salary range, keep it under 200 words,
use plain everyday language.
Do not: use buzzwords, list more than five responsibilities,
or call us a "fast-paced family."

Shines when: you keep getting output that is technically fine but wrong in spirit.

6. Step-by-step reasoning

Ask it to work through the problem before answering. This makes a real difference with anything involving logic, math, or multi-part decisions.

A client project needs 120 hours of work. We have 3 people
available at 60 percent capacity for the next 4 weeks.
Think through the scheduling step by step, then tell me
whether we can deliver on time and where the risk is.

Shines when: the answer requires reasoning, not just writing.

7. Audience specification

Name exactly who will read the output. The same content lands completely differently for a beginner and an expert.

Explain what a cash flow forecast is for someone who just
started their first business and has never seen one.
Use one concrete example with small numbers.

Shines when: you are writing for a specific reader and generic tone will not do.

8. Iterate, don’t restart

Treat the first output as a draft and reply with precise fixes. This beats writing one giant perfect prompt.

Good start. Now make the opening paragraph half as long,
keep the pricing section exactly as is, and make the
closing line more direct.

Shines when: the first draft is 70 percent right and you want 95.

9. Ask for options

Request several variants instead of one. Choosing among five is faster and better than accepting the first.

Give me five subject lines for a newsletter announcing
our new workshop. Mix straightforward and playful tones.
Keep each under 50 characters.

Shines when: quality is subjective and you want to pick, not settle.

10. Critique request

Ask it to attack your work. An assistant playing critic finds holes you are too close to see.

Here is my plan for launching our online course:

[paste plan]

Poke holes in it. What are the three biggest risks, and what would you change?

Shines when: you need honesty before you commit, not encouragement after.

11. Template fill

Provide the skeleton and let it supply the content. You keep full control of structure.

Fill in this project brief template:

Project name:
Goal (one sentence):
Three deliverables:
Timeline:
Biggest risk:
Keep each field to one or two lines.

Shines when: you need consistency across many similar documents.

12. Summarize plus extract

Don’t just ask for a summary. Name the exact pieces to pull out.

Summarize this article in five bullet points. Then list:
1. Every action item mentioned
2. Every deadline or date
3. The one idea worth stealing

[paste article]

Shines when: you need usable output from long material, not a book report.

13. Role reversal

Make the assistant interview you. It asks, you answer, then it drafts from complete information.

I need to write a difficult feedback email to a team member.
Ask me one question at a time until you have enough context
to draft it. Then write the draft.

Shines when: you are not sure what context the task needs.

14. Comparison table

Ask for a side-by-side table when deciding between options. Structured comparison beats two separate paragraphs.

Compare hiring a freelancer versus a part-time employee
for our social media. Table format with rows for cost,
control, speed, and risk. End with your recommendation
and one sentence of reasoning.

Shines when: you are deciding and want the trade-offs visible at a glance.

15. Tone transplant

Keep the content, change the voice. Same facts, different reader.

Rewrite this announcement to sound warm and personal,
like a note from a founder, not a press release.
Keep all the facts and dates exactly the same.

[paste announcement]

Shines when: the content is right but the voice is wrong.

Woman thoughtfully reviewing AI output on a laptop in a modern home office
Read every AI output before you use it; judgment is the real skill.

How do I remember all of these?

Nobody memorizes fifteen patterns, so don’t try. Start with four: context dump, format specification, few-shot examples, and iterate don’t restart. Those four handle the large majority of everyday tasks. Add the rest as situations call for them.

And keep a personal library. Every time a prompt works well, save it where you will find it again. A notes app, a document, or a shared team page all work. Your future self will thank you weekly. If you want a head start, our free Prompt Library has 40 copy-ready prompts with usage notes. They are organized by job: drafting, summarizing, planning, brainstorming, customer replies, and study.

Colorful sticky notes and a notebook of sketched prompt frameworks on a wooden desk
Keep a living library of the prompts that work; your future self will thank you.

Frequently asked questions

Do these patterns work on every AI assistant?

The concepts transfer across all major assistants. Wording may need small tweaks: some handle long pasted context better than others, and a few have quirks with very strict formatting. The pattern matters far more than the platform. Learn the pattern once and you can use it anywhere.

Why do my prompts still give generic answers?

Almost always one of two causes: not enough context pasted in, or no example of what good looks like. Add one of those and try again before you change anything else. Generic input gets generic output, so fix the input first.

Should my prompts be long and detailed?

Long enough to include the context and constraints the task needs, and no longer. A focused 80-word prompt with a good example beats a rambling 400-word prompt every time. When in doubt, add context and cut the throat-clearing at the start.

Can good prompting replace real training?

No, it cannot. Prompts are vocabulary; judgment is the language. Training teaches you which tasks deserve AI at all. It teaches you how to check outputs and build workflows around these tools. Patterns get you better answers today. Training makes you dangerous in the best way.

How do I organize prompts that work?

Save the exact wording and note what task it served. Add one line on what you would change next time. Review the collection monthly and delete what you have outgrown. A living library of thirty good prompts beats a folder of three hundred forgotten ones.

Turn patterns into skill

These patterns will improve your results starting today. But prompting tricks alone are not real AI capability. Capability is judgment. It means knowing which tasks deserve AI at all, and how to check the outputs. It means building workflows that multiply what you can do. That is what our AI training builds, in live hands-on sessions around your actual work.

Start with a 30-minute call and tell us what you are trying to get better at. 30 minutes · No commitment.

Book a call