Your ChatGPT "Mega Prompts" now make your outputs worse.

You know the feeling.

Over time, your super ChatGPT prompts have gotten…(very) long.

You added a rule when an output went sideways. Then another rule. Then another.

Now your “write a blog intro” prompt is 400 words and you copy-paste it everywhere because you’re scared to touch it.

Last month, a head of marketing from one of the largest construction groups in Europe showed up in my workshop, proudly sharing his “mega-prompt” to generate a brief. Months of refinement. Stacked on top of each other. Half date back to GPT-4.

The issue: ChatGPT 5.5 doesn’t need a long prompt anymore.

In a lot of cases, it’s making your output worse.

ChatGPT 5.5 went from junior intern to freelance agency

This 👆🏻 is basically the shift, summarized in plain words.

When you brief a junior intern, you hold their hand, and give them step-by-step instructions:

• Step 1: read this.
• Step 2: identify the themes.
• Step 3: write a first draft.
• Step 4: check the tone| Step 5…| Step 6…| Step 7…

That’s how you train someone who doesn’t know what they’re doing yet, and that’s necessary in their first month.

When you brief a freelance agency, you don’t do that.

You give them a brief: goal, audience, deliverable, must-haves, what to avoid, how you’ll measure success.

Then you let them work.

You review the output, give notes, iterate.

And that’s how ChatGPT 5.5 needs to be briefed now.

It reasons faster, stays direct by default, and makes better choices when you describe the destination instead of every turn along the way.

If you want to train your teams to “actually” use AI efficiently for work, without fluff theory

and learn hands-on to make the best of it,

that’s the core of what we do at Spicy Advisory.

Book a free diagnosis

OpenAI’s own guidance is blunt:

shorter prompts beat long instructions, outcome-first beats step-by-step.

This is relevant, because every model release moves further in that direction. As models improve, they reward cleaner briefs.

The skill that compounds is the brief.

Marketing teams who get it now will be ahead six months from now.

The PROMPT framework

To make it easy to understand, I built a simple framework for marketing, sales, and HR teams who don’t want to learn prompt engineering. They just want their ChatGPT to give good results.

Six parts:

• P = Purpose.

  The outcome you want. Not the activity.

• R = Role.

  The perspective ChatGPT takes.

• O = Output.

  The shape of the answer (table, paragraphs, bullet list…).

• M = Must-follow rules.

  The constraints that genuinely matter.

• P = Proof.

  What evidence to use, what to flag as uncertain.

• T = Test.

  The model checks its own work before answering.

That’s the whole thing. Memorize the 6 letters.

Within a week, your prompts will look more like the briefs you’d send a freelance agency.

Below I share different real-world examples of how to rewrite your prompts to match this new approach

Example 1: Campaign brief

Here is an example of the rewrites for marketing teams I see in our workshops:

Old prompt (the marketing classic):

(Then another 200 words of rules.)

New prompt (P.R.O.M.P.T. framework):

Same job. 80% fewer words.

Better output, because the model picks the path instead of fighting your micro-steps.

The shift is from training the intern to briefing the agency.

The content below is for our Vibe subscribers.

Join us and get:

• Example 2: The persona research prompt rewrite.
• Example 3: The LinkedIn post prompt rewrite
• The trap I see in workshops and the free one-line fix that almost nobody we teach uses.
• The full PROMPT guide on Notion: 4 templates, the audit checklist, the strategy review prompt I use with C-suite clients.
• And full access to all the Vibe resources and articles.

  If you've ever asked ChatGPT to write a persona doc and got a generic B2B SaaS template back instead of your actual buyer, that's the first thing waiting below.

Example 2: Persona research from customer interviews

Old prompt:

New P.R.O.M.P.T.: