10 ChatGPT Prompt Tips That Actually Work
Most ChatGPT users get mediocre results not because the model is bad — but because the prompt is too vague. These 10 techniques cover the most common gaps. Each one has a before/after example you can use immediately.
1. Give ChatGPT a role
Before
“Explain machine learning”
After
Act as a data scientist explaining machine learning to a product manager with no technical background. Use plain language and one concrete analogy.
A role anchors the response. Without one, ChatGPT defaults to a generic encyclopedic tone that rarely matches what you need.
2. Specify the output format
Before
“List ways to improve our onboarding”
After
List 5 ways to improve our SaaS onboarding. Format as: [Problem] → [Fix] → [Expected result]. Keep each row under 30 words.
If you don't specify format, you get prose. Prose is hard to act on. A structured format forces the model to be specific.
3. Set a word or length limit
Before
“Write a summary of this article”
After
Summarize this article in exactly 3 sentences. Each sentence should cover: (1) what the study measured, (2) what it found, (3) what it means for practitioners.
Unbounded prompts get padding. A hard limit forces prioritization and cuts fluff.
4. Provide context about your audience
Before
“Write a product description for our software”
After
Write a product description for our project management software. The audience is solo freelancers aged 25–40 who are overwhelmed by client work. Tone: calm, direct, no corporate speak.
ChatGPT writes for a ghost audience unless you specify one. Define who will read the output.
5. Show an example of what you want
Before
“Write a tweet about our new feature”
After
Write a tweet about our new feature. Style example: "We just shipped something we've been building for 3 months. It's small. It will save you an hour a week. Check it out →" — match that casual, direct tone.
One example is worth a thousand adjectives. It removes ambiguity about tone, length, and style instantly.
6. Tell it what to avoid
Before
“Write an email asking for a meeting”
After
Write a short email requesting a 20-minute call with a potential investor. Do not use phrases like "I hope this finds you well", "synergy", or "quick call". Keep it under 100 words.
Exclusions are as powerful as inclusions. If there's a cliché or format you hate, ban it explicitly.
7. Ask it to think step by step
Before
“Should I raise prices for my SaaS?”
After
Think step by step: what are the key factors I should consider before raising prices for a B2B SaaS with 200 paying customers? After the analysis, give me a clear recommendation.
Chain-of-thought prompting significantly improves reasoning quality on complex decisions. It forces the model to surface its logic before concluding.
8. Separate the task from the constraints
Before
“Write a blog intro about remote work that is engaging and not too long and mentions productivity”
After
Task: Write an opening paragraph for a blog post about remote work. Constraints: Under 80 words. Lead with a surprising statistic. Mention productivity in the second sentence. No clichés.
Mixing task and constraints in one sentence creates confusion. Split them and you get cleaner, more predictable output.
9. Iterate, don't restart
Before
“[Writes a whole new prompt from scratch]”
After
"The tone is right but the second paragraph is too long — cut it in half and add a concrete example where it ends."
Most people abandon a good response instead of refining it. A surgical follow-up prompt is almost always better than starting over.
10. Test your prompt on a real task
Before
“Assume the first response is good enough”
After
Run the same prompt 3 times and compare. If the outputs vary wildly, your prompt is under-specified. If they're all mediocre, one of the dimensions — role, context, format — is missing.
Consistency is the signal. A well-structured prompt produces reliably good output, not occasionally good output.
Test these tips on your own prompts
Paste any prompt into Deepclario and get a score across all 5 dimensions — goal clarity, context, format, constraints, and examples. See exactly which of these tips applies to your prompt, and get a rewritten version that scores higher.