3 prompting techniques I use every day (simpler than they sound)

01 — The problem

Why don't saved prompts improve your results?

You watch AI reels, save prompts, read expert posts. And still, the AI regularly gives you answers you have to redo.

The catch: someone else's prompt is an answer to someone else's task. You find a "sales funnel prompt," paste it into ChatGPT — and it dutifully produces a funnel plan. Too generic to actually work from. Pretty and useless.

What works isn't a prompt collection but task-setting techniques — there are few of them, they transfer to any task, and with them you can assemble the right prompt in a minute yourself. I use the three below every day; in my focus groups we go deeper, but these are the workhorses anyone can handle.

02 — Technique 1

Example prompting: why does showing beat describing?

The simplest and most underrated technique: instead of describing the result you want — show examples of it. Once, on a focus-group call, I opened a clean chat, dropped in reference headlines and asked for more in the same style on a different topic. A minute later I had a pile of headlines ready to use.

A hack inside the hack: references can come from any niche. Take a top fitness headline — "How I lost 15 kg in 3 months without diets or starving" — and it becomes a used-car one: "How I found a Toyota Camry $4,000 below market without brokers or resellers." Viral structure transfers across niches. Build your own example prompt:

03 — Technique 2

Analysis prompting: what does "think first, then do" change?

Instead of "do X," you say: "first analyze A, B and C — then do X." The model gathers information, looks for patterns, and only then produces the result — grounded in its own analysis rather than pulled from thin air.

Here's what it looks like on a real task — a sales email sequence:

Before writing a sales email sequence for (your topic):
1. Analyze the main customer objections.
2. Identify the 5 key psychological triggers that raise
   purchase conversion.
3. Research the optimal sequence structure for a cold
   audience, given the engagement funnel.
4. Find the 3 most effective types of social proof for
   overcoming skepticism.

Based on this analysis, create a plan for a 7-email
sequence — with subject lines, key points and the
psychological trigger for each email.

The pattern transfers to anything: "before drafting a content plan — analyze the audience's pains," "before proposing a landing-page structure — assess the visitor's awareness level." The analysis points are exactly the questions a good specialist would ask before working. You're simply making the model walk that path instead of jumping over it.

04 — Technique 3

Reverse role prompting: why make the AI your critic?

Everyone knows "you are an expert, write X." Far more valuable is flipping it: cast the model not as your assistant but as your client — and ask for honest criticism. An example for a fitness app:

You are a young mother on maternity leave with an 8-month-old.
You want to get back in shape, but you have very little time,
you are constantly tired, and you can only work out at home
while the baby sleeps.

Look at this fitness app description and tell me honestly:
1. What you like about it and find useful.
2. What triggers skepticism or annoys you.
3. Which features you are missing.
4. What you would pay for — and what seems useless.

The same move for other tasks:

For pitch decks: "You are an investor who has heard hundreds
of pitches. What annoys you in startup decks — and what
actually catches your attention?"

For content: "You are my typical reader: a man of 35–45,
a small business owner who does not trust marketers. Read
this text and tell me honestly what annoys you and what
hooked you."

The more specific the role — age, circumstances, tiredness, skepticism — the more honest and useful the critique. "Rate my text" gets you compliments. The exhausted-mother role gets you the truth.

05 — Combining

How do the techniques stack into a chain that actually works?

The biggest mistake in working with AI is scattered generic prompts. One "build me a funnel" prompt yields a pretty, non-working plan. A working solution is a chain, where each step's output feeds the next:

→ Give the model your product info (example: show what exists)
→ Analyze the customer's thinking by awareness level (analysis)
→ Shape the product's meanings and value (based on the analysis)
→ Stress-test the offer through a skeptical client (role-critic)
→ Build the lead magnet → write the warm-up → assemble the funnel

See how the three techniques slot into one process? Example gives the model your context, analysis makes it think before producing, and the role-critic checks the result before a live audience ever sees it. That's no longer a "prompt from a reel" — it's an algorithm for solving the task, and the output is night-and-day different.

1. "Here are examples of what I like: [references]" (example)
2. "Before doing anything, analyze [audience/objections/
   structure]" (analysis)
3. "Do [the task] based on the analysis"
4. "Now become my client [specific role] and critique the
   result" (role-critic)
5. "Fix it based on the critique"

06 — Choosing

Which technique should you start with?

If you're unsure where to start — answer three questions about the task currently on your plate: