Best Practices
Below is a recommended set of best practices to help you learn prompt writing faster and get stronger outputs when using generative AI. This list is meant to be helpful, not complete.
Adjust Your Output
When results aren’t what you expected, try adjusting:
Tone (formal, casual, friendly, academic, etc.)
Length (short summary vs. detailed explanation)
Voice (enthusiastic, empathetic, optimistic, authoritative, etc.)
Avoid words that could be interpreted multiple ways when possible.
Ask for multiple versions or alternative outputs if the first response misses the mark.
Be Specific
Clear, detailed prompts usually produce better results.
Ask the AI to ask you follow-up questions if information is missing or unclear.
If you’re creating web/Canvas content, request the output in a specific format like table / markdown / HTML.
If the response feels vague, add more context or constraints.
Provide examples of what “good” looks like.
Role prompting can also reduce confusion and make outputs more relevant. For example:
“Act as an expert in ___.”
Examples: Copywriter, Public Speaker, Marketing Strategist, UX Designer, SEO Expert, Programming Tutor, Web Development Expert.
Set Boundaries
You can guide the AI by stating what to include and what to avoid.
Point out mistakes and ask it to correct them.
Tell it what you do / don’t like about the output.
Ask it to stay focused on the topic.
Repeat key instructions if needed.
Define what it should and should not cover.
Prompting Strategies
If your task is complex, try breaking it into steps:
Use sequential prompting: tell the AI you’ll provide more input next.
Ask it to confirm when it’s ready for the next piece.
If the output is off, pause and rephrase (e.g., “Make it shorter and less formal.”)
Request reasoning when helpful: “Why did you choose that approach?”
Prompt Examples & Libraries
For additional guidance, you can review example prompts that have produced strong results and use them as inspiration. Prompt libraries can also provide templates for common use cases.
Examples (Why specificity matters)
Be specific
A prompt like “Write a story” often leads to a generic story. You’ll get better results by specifying genre, audience, tone, length, and key details.
Specificity can also reduce inaccurate answers. For example, asking “What’s the best restaurant in York?” might return results for York in the UK instead of Ontario. Asking “What’s the best restaurant in Yorkville, Toronto, within walking distance of the Royal Ontario Museum?” gives the AI a clearer target.
“Act as if…”
You can ask the AI to respond from a role or perspective to better match your needs.
Example:
Instead of “Create a recipe using these ingredients,” try:
“Act as my personal trainer. Create a recipe using these ingredients that helps with post-workout recovery.”
Tell it how to present the result
Generative AI can produce many formats. If you want a specific structure, say so.
Try phrasing like:
“Present this as a checklist.”
“Write this as a short email.”
“Return the result in a table.”
“Create an outline with headings and bullets.”
Use “do” and “don’t”
Clear instructions about what to include or avoid can save time.
Example:
Help me write a thesis statement for my essay on climate change.
Do: Create one clear, arguable thesis statement for a 5-page essay, focusing on how human activity contributes to global warming.
Don’t: Write a broad, generic statement or include unrelated facts about the environment.
Use examples (carefully)
Providing a sample sentence or style reference can improve results.
Avoid using copyrighted materials or asking the AI to reproduce something closely.
Consider tone and audience
Specify who it’s for and how it should sound.
Example:
“Give me ideas for a best man speech that’s funny and heartfelt, but appropriate for a family audience.”
Build on previous prompts
You don’t need the perfect prompt on the first try. Start simple, then refine by adding more context, constraints, or clarifying instructions.
Correct mistakes and give feedback
Treat the AI like a teammate. Tell it what worked, what didn’t, and ask it to revise. If it gets something wrong, point it out and request a correction.
Ask the AI what it needs from you
If you’re stuck, you can ask the AI to help you craft the prompt or identify missing information.
Examples:
“Help me write a prompt to accomplish this task: ___.”
“What information do you need to do this well?”
Sources
Adapted for Kritik based on guidance from University of Michigan and Harvard University IT.