How to Write Your First AI Prompt
Written by the AIVX Labs team · Published August 2026
A prompt is just an instruction, and like any instruction it works better with more relevant detail and worse with less. This walks through three real before/after examples so the difference between a vague prompt and a specific one is concrete, not abstract advice to “be more specific.”
Key takeaways
- A vague prompt forces the AI to guess at context you actually know — a specific prompt hands that context over instead.
- The four things worth adding to almost any prompt are: purpose, audience, format/length, and tone.
- Pasting a real example of what you want (an old email, a past post) does more for output quality than almost any other single change.
- Your first draft of a prompt doesn't need to be perfect — refining it after seeing the output is a normal, expected part of the process.
Why vague prompts fail
The AI Beginner Courses channel on YouTube has several videos built around exactly this before/after comparison, if seeing it happen live helps. A vague prompt gives the AI nothing to differentiate your request from millions of similar ones, so it produces the statistically generic version — safe, broad, and usually not what you actually needed. It's not that the AI is incapable of a good answer; it's that a vague prompt hasn't told it which of many reasonable answers you want.
Example 1: an email
Vague
“Write an email to my team about the deadline.”
Specific
“Write a short email to my 5-person design team letting them know the client deadline moved from Friday to next Wednesday. Keep it under 100 words, calm and matter-of-fact (not apologetic), and end with one clear next step: send me your updated timelines by end of day tomorrow.”
The vague version doesn't say which deadline, who the team is, how the news should land emotionally, or what the reader should do next — so the AI invents plausible-sounding answers to all four questions, and the odds they match what you had in mind are low.
Example 2: a summary
Vague
“Summarize this article.”
Specific
“Summarize this article in 4 bullet points, focused specifically on what it says about pricing strategy — I don't need the background or the author's personal anecdotes, just the pricing-relevant takeaways. [paste article]”
“Summarize this” alone leaves the AI to guess what “summary” means to you — length, focus, format all unspecified. Naming the exact angle you care about (pricing strategy, in this case) filters out everything you'd otherwise have to skim past yourself.
Example 3: a social post
Vague
“Write a LinkedIn post about our product launch.”
Specific
“Write a LinkedIn post announcing our new scheduling feature launched this week. Audience: small business owners who already follow us. Tone: excited but not salesy. 3-4 short paragraphs, end with a question to drive comments, not a hard CTA. Here's our last launch post as a style reference: [paste]”
The four additions that do the most work
Across almost any task, four additions consistently improve a vague prompt more than anything else: purpose (what this is actually for), audience (who will read or use it), format and length (how long, what structure), and tone (formal, casual, urgent, warm). Add those four to almost any request and the output moves from generic to genuinely usable.
Want to see this practiced live?
The AI Beginner Courses channel on YouTube walks through building prompts like these from scratch, on real tasks.
Refining after the first response
Even a well-built prompt often needs one round of refinement — that's normal, not a sign the prompt failed. Rather than rewriting the whole thing, respond in the same conversation with a specific correction: “good, but cut this to half the length” or “too formal, make it sound more like me.” This describe-review-refine loop is most of what separates people who find AI useful from people who give up after one mediocre response — see common AI beginner mistakes for more on that specific failure pattern, or What Is Prompt Engineering? for the more advanced version of this same idea.
Read our featured article on LinkedIn
Every course and tool mentioned here is included free on AIVX Labs.
Create Your Account NowFrequently asked questions
What makes an AI prompt good instead of vague?
A good prompt gives the AI the same context you'd give a person doing the task for you: what it's for, who it's for, how long or what format, and the tone you want. A vague prompt skips all of that and forces the AI to guess, which produces a generic result that usually needs heavy editing.
Do I need special phrasing or keywords to write a good AI prompt?
No — there's no secret syntax or magic phrase required. Writing in plain, clear language the way you'd brief a coworker works better than trying to sound technical or use unfamiliar prompt-engineering jargon you're not sure you're using correctly.
Can I fix a bad AI response instead of rewriting my whole prompt?
Yes, and this is usually faster than starting over. Say exactly what's wrong with the response — too long, wrong tone, missing a specific point — and ask for a revision within the same conversation, since the AI still has the original context.
What's the single most useful thing to add to a prompt?
A real example of what you want, if you have one — a past email, a previous post, a sample of the format. Pasting an actual reference does more to align the output with what you have in mind than almost any amount of additional written instruction.