What if I were to tell you that I know why your experience with AI hasn’t matched the hype?

If you’ve been scrolling through LinkedIn, wondering why everyone else seems to be having life-changing breakthroughs with AI while your own experience feels more… meh, you’re not alone.

Some of that is just the overblown hype of folks gaming the LinkedIn algorithm. “I tried ChatGPT-5” doesn’t get as many views as “ChatGPT-5 changed my life, and it’ll change yours too.” That said, there’s still good reason for people to feel the way they feel.

It’s not that ChatGPT is underwhelming. It’s not that you’re bad at writing prompts. And it’s not that the people posting screenshots are superhuman prompt whisperers. The real problem is the communication gap between you and the model. To put it simply: you’re not speaking the same language.

AI speaks in tokens. You speak in an actual human language. What you write has to be translated into a series of alphanumeric strings that the AI then parses.

To bridge that gap, try using the RCTO framework.

RCTO helps you get the most out of AI

RCTO stands for “Role, Context, Task, and Output.”

Think of it as compressing a creative brief into four key beats:

• Role: Tell the model who it should be.

• Context: Explain the project background and boundaries.

• Task: Define exactly what you want the AI to do.

• Output: Specify the form, format, and tone the answer should take.

That’s it. Four switches that move your results from undesirable to professional-grade. 

If you give AI unstructured instructions in a creative stream of consciousness, the AI will typically generate messy and unstructured answers. But if you package your ask through RCTO, the model has everything it needs to deliver something you’d actually consider moving forward with.

Of course, it’s important to remember that AI outputs are rough drafts only and still need to undergo creative review for quality and strategic alignment.

Role does not just mean job title

When you set a role, you’re not just assigning a function. You’re giving the model a perspective. Instead of “be a copywriter,” try saying something like this: “You’re a creative director critiquing ad headlines for voice consistency.” The nuance shapes the lens through which it answers.

You could also consider projecting. If you don’t know what role to provide it, just give it your role—since you’re having it help you with your work. I could’ve told the AI helping me write this blog article to “assume the role of a brand strategist with expertise in AI prompting for creative services.” You can also just ask the AI itself, through a process called metaprompting.

Context is not a data dump

Dumping paragraphs of background into your prompt doesn’t help it. AI models are easily overwhelmed when you provide too much information. Context should be curated—if it doesn’t belong on the first slide of a brief, it probably doesn’t belong in your prompt.

Provide the essentials: like the background, constraints, and guiding brand principles. That’s it.

Tasks need to be specific

“Write webpage copy” is too vague. Prefer actionable copy instead, like so: “Draft three alternate H1s for the hero section of a D2C product launch landing page targeting eco-conscious Gen-Z customers.”

Specificity accelerates understanding. Understanding accelerates iteration. And, in the end, you claw back some of the would-be wasted time spent on going back and forth answering clarifying questions.

Output should be treated like a contract

The most overlooked part of prompting is telling the model exactly how you want the result to be delivered. Output should include format (e.g., bullets, paragraphs, table), length, tone, and style guidelines. It’s not just “what” you want—it’s the shape of the deliverable.

Otherwise, the model will probably just pick whichever output it guesses most aligns with your goals. That tends to mean using an excessive amount of bullet points and em dashes.

Before and after: rambling prompts vs RCTO prompts

Here’s a quick side-by-side comparison:

The rambling prompt will get you something you might have a use for at some time. The RCTO prompt will get you crisp, differentiated headlines ready to share with your creative team for refinement.

RCTO opens the door to optional add-ons

The ABCs of AI prompting might start with RCTO, but it doesn’t end there.

E: Examples

Golden samples are powerful. Show the model what “great” looks like and be sure to tell it how rigidly to reflect them (like an opacity slider in Figma). You can tell it to loosely emulate the style, or to follow the example as a strict template and only swap out subject matter.

R (again): Reasoning effort

With models like GPT-5, you can ask the model to think harder or to answer faster. To ask for speed, use natural language that reflects a low reasoning parameter, like “give me a fast draft.” For high effort, ask the AI to “think longer for a better answer” or “explore multiple options and refine before answering.”

If you’re working in the GPT-5 API, reasoning_effort is an explicit parameter, which makes it even easier to tune.

V: Verbosity

Verbosity controls the length of the answer. GPT-5 respects natural language overrides, so prompts like “Keep it under 150 words” or “Provide a comprehensive, more detailed explanation” work well.

If you’re still not happy with AI prompting, try metaprompting

Sometimes, you use RCTO perfectly—or think you use it perfectly—and the output still misses the mark. That’s when you hand control back to the model in a process called metaprompting.

When you need to improve your prompt, just ask the AI, “What could I add to or remove from this prompt to better achieve [X behavior or result]?”

Because GPT-5 was trained to reflect on instructions, it’s surprisingly good at helping you fix your own prompt design. And other leading models are now being trained to do the same.

Case study: Cursor and GPT-5

One of the more interesting case studies on AI prompting comes from Cursor, an AI-powered code editor. Their team found that older models needed heavy encouragement to analyze context, so they loaded their prompts with “maximize context understanding” instructions.

But when they tried the same with GPT-5, it backfired. The model was already naturally proactive, so the extra wording caused over-searching and wasted tool calls. By simplifying their prompts—removing the “maximize” language and tightening the structure—they got better results: cleaner code, fewer delays, and smoother long-horizon tasks.

RCTO worlds because it prioritizes clarity and treats verbosity as a tool for achieving that clarity. More words don’t necessarily make a better prompt, but using the RCTO structure almost always does.

Key takeaways from these AI prompting tips

These AI prompting tips show you exactly how to apply the RCTO framework for clearer, more effective results:

• Use RCTO as your base. Role, Context, Task, and Output transform vague instructions into clear prompts that deliver predictable, brand-ready results.
  
  

• Define Role as perspective. Go beyond job titles. Assign the model a viewpoint, like “creative director reviewing ad copy,” to shape tone and direction.
  
  

• Curate Context carefully. Provide only the essential background, constraints, and brand guidelines. Focused context keeps the model on track and prevents irrelevant tangents.
  
  

• Make the Task specific. Replace broad asks with precise actions, such as “Write three landing page headlines targeting eco-conscious millennials under eight words each.”
  
  

• Treat Output as a contract. Specify structure, tone, and length. Example: “Deliver in bullet points, playful tone, under 150 words.”
  
  

• Add optional enhancers. Use examples to anchor style, reasoning effort for depth vs. speed, and verbosity to control response length.
  
  

• Use metaprompting to refine. Ask the model what to change in your prompt when results miss the mark. GPT-5 adapts well to this technique.
  
  

• Maintain a prompt library. Archive winning prompts, retire weak ones, and categorize by use case. A database of prompts is today’s creative style guide.

Maintain your prompt library like a style guide

The best prompts aren’t one-offs. They’re reusable assets. So if you find yourself spending half an hour writing a prompt for a task that would’ve taken half that time if you just did it yourself, that’s fine. Next time, you’ll do it in seconds.

Just like a brand style guide holds onto logos, fonts, and approved phrases, you should maintain a database of prompts that worked. And be sure to annotate that database with notes about performance.

Rotate out stale examples. Archive your wins. Tag prompts by use case. Treat this database like your new creative ops backbone—because that’s exactly what it is.

With RCTO, you’ll create prompts that are measurable, repeatable, and improvable. And in the end, you’ll find yourself transforming AI from a novelty tool into a true extension of your own techno-creative process.

Learn more about Hunter Amato’s approach to techno-creative ops.