Seven prompting habits that instantly improve your AI output
Prompt engineering isn't a secret vocabulary. It's seven habits, and you can pick all of them up in one session.
1. Say who it's for and what happens next
The single biggest quality jump comes from telling the model the audience and the destination. "Write a project update" and "write a project update for a non-technical client who is worried about the timeline, to be sent as an email today" produce completely different results from the same model.
2. Give an example of good
One example of the output you want beats three paragraphs describing it. Paste a past piece you were happy with and say "match this tone and structure." This is the closest thing to a cheat code in everyday AI use, and almost nobody does it.
3. Constrain the shape
Length, format and structure should be explicit. Five bullets. Under 150 words. Three sections with these headings. A table with these columns. Without constraints the model picks its own, and its defaults are longer and blander than what you wanted.
4. Ask for options, not an answer
"Give me five different angles, one line each" is far more useful than "write it". You pick the angle — the thing you're actually better at than the model — and then ask it to develop the one you chose. This mirrors how good work happens with a human collaborator too.
5. Iterate in place instead of restarting
When output is 70% right, don't rewrite the prompt from scratch. Say what to change: "same structure, but cut the intro and make the tone more direct." Restarting throws away the context that got you to 70%.
6. Make it ask you questions
End your prompt with "ask me up to three questions before you start." You'll get questions about details you forgot to specify, and the output after answering them is consistently better than the output you'd have got from a longer prompt.
7. Keep the prompts that work
When a prompt produces something great, save it. A personal library of ten proven prompts for the things you do weekly is worth more than any collection of a thousand generic templates, because it's built around your actual work.
Tools mentioned
Frequently asked questions
Do prompting techniques differ between tools?+
The fundamentals — context, examples, constraints, iteration — transfer everywhere. Tool-specific tricks exist but they're a small fraction of the quality difference.
Should I learn prompt engineering formally?+
No course needed for everyday use. These seven habits cover the large majority of the gain; the rest is practice on your own work.
Find the tools that fit your situation
Take the 30 second quiz and get a personalized shortlist — free.
Get my personalized shortlist