← Back to Blog
Prompt Engineering · 4 min read

5 Prompting Best Practices That Actually Work

I used to send vague prompts like "write a blog post" and get back generic fluff. Then I tried adding who I am and who it's for — "As a marketing director, write 500 words for small business owners on how AI saves time" — and the output was usable on the first try. Context matters more than I expected.

Another thing: I always add "keep it under 15 words per bullet" or "3 bullet points max" when I need slides. Without that, ChatGPT gives me paragraphs. Break big asks into steps, say what format you want, and don't be afraid to ask again if it misses. Your first prompt is a draft, not the final answer.

1. Add context up front

Tell the AI who you are, who you're writing for, and why it matters. "As a PM, write a feature spec for engineers" beats "write a spec" every time.

2. Specify format

Bullets, word counts, section headers — say what you want. "5 bullets, 15 words each" or "3 paragraphs, 100 words total" saves a lot of editing.

3. Break it into steps

For complex tasks, list steps: "Step 1: summarize the data. Step 2: identify 3 trends. Step 3: write 2 recommendations." Keeps the AI from wandering.

4. Add constraints

"Don't guess at causes." "No jargon." "Grade-8 reading level." Constraints focus the output and cut the fluff.

5. Iterate

Your first prompt rarely nails it. Ask for edits: "Shorter." "More casual." "Add an example." Treat it like a first draft.

Explore our prompt library →