The prompt formula I ignored

For a year I blamed the tool for my own lazy prompts

July 10, 2026

Sponsored by

I'll admit something a little embarrassing. For a long time I typed throwaway questions into ChatGPT and expected genius back. "Tell me about marketing." That kind of thing. The answers came back flat, generic, forgettable, and I blamed the model every time.

Then I read a breakdown from an AI professional that quietly flipped how I think about all of this. I wish I'd read it a year earlier. It would have saved me hundreds of mediocre answers and a lot of misplaced frustration.

The person who wrote it made the exact same mistake when they first got serious about AI. Vague question, wait for magic, get back filler they could have Googled in 30 seconds. So they changed one habit, and everything shifted.

How Jennifer Aniston’s LolaVie brand grew sales 40% with CTV ads

The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.

*Ad

The one shift: stop writing search queries, start writing briefs

Here's the mental model, and it stuck with me instantly. Think about how you'd brief a new hire before handing off a project. You'd give them the goal, the background, the constraints, and a clear picture of what "done" looks like. You wouldn't say "do marketing" and walk out of the room.

That's the whole move. Treat the prompt like a brief, not a Google search. The tool was never the problem. The prompt was. And the jump in output quality is immediate the second you start writing this way.

Then they broke down what a genuinely great prompt is actually made of. It isn't one magic sentence. It's 10 components, each carrying its own weight. What I liked most is that each piece got a rough percentage, so you can see how much it actually moves the needle.

The best voice models, now fully orchestrated across all channels

ElevenAgents puts full orchestration on the voice models the market builds around. Voice, text chat, transcription, and reasoning in one integrated stack, <400 milliseconds, and human-sounding. Plug in any LLM, integrate tools, A/B test, and deploy across channels. More human-like conversations, lower latency, flat $0.08 per minute.

*Ad

The 10 building blocks, ranked by how much they matter

  1. Context (20%). The heaviest piece, and it makes sense. Hand over your business type, industry, audience, goals, and constraints. More context almost always means better output. Don't make the model guess your situation.

  2. Objective (15%). Define the task with zero fog. "Tell me about marketing" gets noise. "Create a 90-day marketing strategy for a SaaS startup targeting small businesses" gets a plan you can actually use. Specificity is the entire game.

  3. Input Data (15%). Give it the raw material. Meeting notes, customer feedback, research reports, whatever's relevant. The rule is blunt and correct: never assume it knows your situation.

  4. Role (10%). Tell the model who to be. Senior software engineer. Expert copywriter. Startup advisor. Each role activates a different reasoning pattern, so the same question gets answered from a sharper angle.

  5. Instructions (10%). Spell out the exact actions. Analyse. Identify. Prioritise. Action verbs do real work, so pick them on purpose instead of leaving the model to interpret.

  6. Constraints (8%). Set the guardrails. Max word count, no jargon, beginner-friendly, a specific budget range. Limits force focus, and focus improves consistency.

  7. Output Format (8%). Tell it the shape you want. Table, bullet points, step-by-step guide, JSON. Models follow structure reliably when you ask for it explicitly.

  8. Examples (5%). Show it what "good" looks like. Give an input, the desired output, and the format. Examples cut misinterpretation down in a big way.

  9. Iteration Request (5%). Ask it to improve on itself. "Critique your response." "Suggest three alternatives." Prompting works best as a back-and-forth, not a one-shot lottery ticket.

  10. Quality Checks (4%). Ask the model to review its own work. Flag weak assumptions. Point out what's missing. This tiny addition catches a surprising number of mistakes before they ever reach you.

You don't need all 10 in every prompt. That's the part I love. But when an answer comes back weak, you now have a checklist to diagnose why. Missing context? No role? No format? You can find the gap in seconds instead of just retyping and hoping.

HR and IT need to work as one. Here's how

HR and IT misalignment costs you more than time. Get the guide that closes the gaps across the full employee lifecycle.

*Ad

Why the structure actually works

Vague prompts leave the model to fill in a hundred blanks, and it fills them with the most average, most predictable stuff it has. That's the "filler" I kept getting and blaming the tool for. Every component you add removes a blank and points the model at your specific problem instead of the generic version of it.

Structure isn't extra effort. It's the difference between a search-bar answer and something you can actually put to work.

Where I'd start tonight

You don't have to memorize the whole framework. The person who shared this started with two moves, and I'd suggest the same, in order:

  1. Add a role and a clear objective to your next prompt. Just those two. Watch how much sharper the reply comes back.

  2. Then layer in context and an output format. Those four alone cover most of the quality jump. This is the 80/20 of the whole thing.

  3. Once that feels natural, use the iteration request to refine instead of restarting. "Critique this and give me a stronger version" beats deleting and starting over.

Build the habit one piece at a time and it becomes automatic. Soon you'll be writing structured briefs without thinking about it, and your "average" AI results quietly disappear.

The gut-check question I'll leave you with is the same one that closed the original post: are you a structured prompt writer, or still winging it? No shame either way. I was firmly in the winging-it camp not long ago. Pick one prompt tonight, add a role and an objective, and feel the difference before you scale it to the rest.