The worlds best prompters use this prompting structure

Hey guys,

Besides copywriting, prompt and context engineering is the most important skill that a creative strategist can learn in 2025.

But most people still suck at prompting.

So today I'm gonna be sharing my 4-step process to creating intentional prompts that help me extract successful outputs from AI.

But first, here’s what’s I’ve come across in AI this week:

• Taylor Holiday dropped a good video on using AI as a manager in orgs

• Dan Koe shared a cool AI workflow that breaks down any skill into a 90 day study regimen

• A Reddit user deposited $400 into Robinhood, then let ChatGPT pick option trades and it had a 100% win rate over 10 days lol

And this isn’t AI, but I loved this tweet by Barry Hott on content relevance in 2025.

The 4 Elements of Every Great Prompt

Here are the 4 core elements that I try to work into the majority of my prompts:

Identity + Task + Context + Output

Identity

This is about telling the AI who it should be.

E.g. “You are a seasoned direct response copywriter trained to extract winning Facebook ad headlines directly from customer reviews. You write like the best performance marketers in the world — people like David Ogilvy, Dave Gerhardt, and Harmon Brothers.”

Task

Clearly define exactly what you want the model to do.

E.g. “Your job is to analyze the customer reviews I provide and select the ones that are most suitable for static Facebook ad headlines. Then, rank these review snippets by how confident you are that each could work as a headline for a prospect who has never heard of our brand. Prioritize relevance, specificity, and emotional clarity over generic praise or brand loyalty.”

Context

Give the AI both good and bad examples, and explain WHY they're good or bad (this is arguably the most important part).

E.g. “A good example:

“The only shampoo that ended my dandruff battle”

— Clear pain point, outcome-driven, universally relatable

A bad example:

“I’m super picky, and these guys hit it out of the park from A to Z.”

— Too vague, not specific to any particular pain point or audience”

Output Structure

Tell the AI EXACTLY how you want the response formatted for a more accurate output:

E.g. “A ranked list of 5–10 customer quotes (best to worst), with each quote labeled A, B, C, etc.

Under each quote, add a short justification: explain why this one is strong/weak for cold audiences.

Then, group them into tiers:

• Tier 1 = Highly usable as-is

• Tier 2 = Might need light editing

• Tier 3 = Too vague”

So altogether, once we’ve added a little more meat to this prompt here’s what it looks like:

A few prompting hacks…

Here are some other lines I like to add to my prompts to help me improve the outputs:

• Rate each output from 1-10 and only show me 8+ quality

• Once you’re happy with your final output, ask "What would I have had to ask to get this exact output?" (reverse engineer your wins)

• For each output, tell it that it was a 4/10, and you need a 10/10 - impress me

• Raise the stakes by adding "if this does not [INSERT GOAL], I will get fired. It is that important"

• Ask “Now rewrite this as if it's version 10 after 9 rounds of expert feedback”

• At the end of your initial prompt, add in this sentence: “Before you give your response, what questions do you have that would help you give a better response?”

• For each output, ask, “How could you improve that?” Keep asking this question over and over

• Tell it to act as an expert. E.g. "Act as a world class direct response copywriter" or "write like you're being paid $1,000,000 to do this"

Raising the stakes has been proven to improve the outputs that AI gives you. Hence why I end up using these all the time haha.

Build Your Prompt Library

And finally, once you have started making prompts like this, you’ll want to share them across your team.

Building a prompt library is something I’d recommend every creative team out there does today.

It just saves so much time in people not rebuilding prompts that have already been built.

TLDR:

• Use the Identity, Task, Context, Output Structure format for your prompts

• Raise the stakes to improve the models’ outputs

• Build prompt libraries to share learnings across your team

Announcements:

We’re now half way through the AI Creative Strategist Blueprint. Last week we had another fun session on prompting - this time for DeepResearch and Image Gen.

If you haven’t joined the program yet, all replays of past sessions are made available so that you can catch up without missing a training! Click here to sign your team up.

Enjoyed this email? Forward it to a friend who’d find it valuable!

And if someone forwarded this email to you, here’s a link to signup. I send tips on how to make more winning ads with the help of AI straight to your inbox every week!

See you next week,

Alex