25 Prompts to Teach ChatGPT-5 Your Thinking Style

Fast-track your AI’s learning curve today.

You don’t get the output you deserve — you get the output you specify.
Most people toss vague questions at ChatGPT-5 and pray. You and I? We’ll install your thinking style so the AI answers like a sharper version of you. Starting now.

Clarity scales. Vague dies.

Your goal is to make ChatGPT-5 think, write, and decide like you — for work, content, and strategy.
The problem is that you keep rewriting AI’s answers because it doesn’t know your voice, priorities, or decision rules.
One key takeaway: teach your style first, then ask for work.

Why This Works (60-Second Mental Model)

• LLMs mirror the context you provide. When you define role, audience, tone, constraints, and examples, quality jumps. (See platform docs on custom instructions and prompting best practices.)
• Iteration beats one-shot asks: small cycles of “draft → critique → revise” outperform single mega-prompts.
• Clear, descriptive headings and concise writing help both readers and search engines. Put your primary keyword early in the title/H1 and echo related phrases in H2/H3. Keep sentences short.

Most ‘AI problems’ are prompt problems.

Note: Features like Custom Instructions and Memory vary by model and plan. Check your tool’s settings to see what’s available.

Setup (2 Minutes)

1. Open a fresh chat.
2. Grab 2–3 short samples of your writing/notes.
3. Paste the prompts below.
4. If your tool supports saved instructions or memory, save the final “Style Card” so every new chat starts with your voice. (ChatGPT offers Custom Instructions; other assistants provide similar controls.)

The 25 Prompts (Copy-Paste Templates)

How to use: Replace the {brackets}. Keep it conversational. Ask for a brief “why” summary — no need for a long chain of thought.

A) Calibrate the compass (identity + audience)

1. Role & Outcome: “Act as my {role}. Your job: {Outcome}. Before answering, restate the goal in one line.”
2. Audience Lens: “Assume my reader is {audience}. Prior knowledge: {low/med/high}. Optimize for {clarity/speed/detail}.”
3. Constraints: “Use {max word count}, {bullets over paragraphs}, and avoid {jargon list}.”
4. Success Criteria: “A+ answer = {3–5 criteria}. Grade your output A–F against them, briefly.”
5. Non-Goals: “Do not {list}. If needed, say ‘I don’t know’ and suggest where to look.”

B) Install your voice (style, tone, cadence)

1. Voice Snapshot: “Here are 2 samples of my writing: {paste}. Extract a style guide (tone, cadence, sentence length, transitions, do/don’ts).”
2. Micro-Style: “Summarize my favorite phrasing habits as 10 rules (‘Use punchy openers’, ‘Kill filler’).”
3. Tone Switcher: “Default tone: {e.g., warm, direct}. If topic = {sensitive}, shift to {tone}.”
4. Formatting DNA: “From now on: short sentences, scannable subheads, bullets > walls of text.”
5. Style Card: “Condense my style into a 120-word ‘Style Card’ I can paste into any AI.”

Voice is structure wearing confidence.

C) Teach your thinking (decision rules)

1. Priorities: “When trade-offs appear, rank in this order: {e.g., accuracy > speed > creativity}. Apply explicitly.”
2. Frameworks: “Default frameworks: {e.g., 80/20, first-principles, Eisenhower}. Show which one you used in one line.”
3. Risk Guardrails: “If claim confidence < 80%, flag it and propose verification steps.”
4. Assumption Check: “List 3 assumptions you made and 1 way to test each.”
5. Bias Brake: “Offer 2 opposing takes before concluding. Keep it to 2 bullets.”

D) Shape the output (usefulness over volume)

1. Answer Shape: “Give me: TL;DR (2 lines) → Why it matters (3 bullets) → Action steps (5 bullets) → Pitfalls (3 bullets).”
2. Examples: “For each idea, include 1 crisp example relevant to {industry}.”
3. Numbers: “Prefer concrete numbers/benchmarks if available; otherwise say ‘no solid data.’”
4. Clarity First: “Rewrite any long sentence into two shorter ones. Prefer verbs. Cut filler.”
5. Medium Formatting: “Use skimmable H2/H3 with keywords near the front; keep hierarchy clean.”

E) Build the feedback loop (improve every reply)

1. Self-Review: “Before final, rate the draft on {criteria}. Give 3 fixes and apply them.”
2. Delta Pass: “Here’s feedback {paste}. Produce a ‘Δ-revision’ showing only changed lines.”
3. Taste Match: “From 1–10, how well does this match my Style Card? What’s missing?”
4. Memory Nudge: “Remember these permanent preferences: {bullets}. Confirm back in 2 lines.” (If your tool supports memory.)
5. Continuous Upgrade: “When you notice my new habits, propose Style-Card updates monthly in 6 bullets.”

Feedback is the fastest way to borrow tomorrow’s skill.

Before/After Demo (What Changes When You “Teach” Your AI)

Before (vague ask):
“Write a post about productivity.”

After (using the system):

“Act as my content strategist. Outcome: a 900-word Medium post. Audience: busy knowledge workers with low tolerance for jargon. Constraints: use short sentences, place H2/H3 keywords early, and prefer bullets over paragraphs. Success: scannable, practical, 3 examples. Non-goals: generic tips. Here are 2 writing samples: {paste}. Extract my Style Card in 120 words, then draft. Use TL;DR → Why it matters → Steps → Pitfalls. Confidence-check claims; if <80%, say so.”

Result: tighter voice, more transparent structure, fewer rewrites.

Mini playbook: Publishing that wins on Medium

• Lead with promise/problem in the first 2–3 lines. That’s what stops the scroll.
• Front-load your primary keyword in the title/H1 and echo related phrases in H2/H3. Write descriptive headings that clearly indicate what I can expect.
• Keep paragraphs short; break complex bits into bullets; end sections with a one-line takeaway.

Common Mistakes (And Quick Fixes)

• Vague asks → Add role, audience, constraints, examples, and success criteria.
• One-shot prompts → Iterate; show a sample; refine with self-review. (
• Overlong answers → Enforce an output shape and word limits. (Your Prompt #D — 1.)
• Assumptions hiding in plain sight → Use the Bias Brake and Assumption Check (Prompts #C — 4 & 5).

Your Style Card (Screenshot-Able)

Copy this once, then tweak.

STYLE CARD — {Your Name}

Voice: {direct, warm, pragmatic}. Cadence: short sentences; active verbs; no filler.

Audience: {who they are, what they want}. Readability target: Grade 6–8.

Do: lead with promise/problem; use H2/H3 with keywords early; show 1 example per idea.

Don’t: hedge, bury the lede, stack long paragraphs, overuse adverbs.

Tone switches: {sensitive topics → calm, validating}.

Priorities: accuracy > clarity > speed > creativity.

Formatting DNA: bullets > walls; bold key phrases; TL;DR → Why → Steps → Pitfalls.

Hallmarks: metaphor-lite; numbers when solid; “I/you” voice; 1-line takeaways.

Voice isn’t a mood — it’s a system.

Resource Box (Save This)

• One-liner to generate your Style Card:
  “Here are 2 samples of my writing: {paste}. Extract a 120-word Style Card (tone, cadence, sentence length, do/do n’ts, formatting DNA). Keep it practical.”
• One-liner to reuse your style everywhere:
  “Use this Style Card {paste} for all replies in this thread unless I say otherwise. Confirm in one line.”
• One-liner to speed-test a draft:
  “Give me a 10-second skim test: what did you learn, what’s missing, where did you slow down?”

Notes on Tools (Brief, so You Don’t Chase Settings)

• ChatGPT: Supports Custom Instructions across plans; you can save preferences that apply to new chats. Memory features are evolving; check Settings → Memory for your account.
• Claude & Gemini: both publish prompt-engineering guides with similar best practices (be specific, set roles, show examples, iterate). Names and UI differ, but the ideas are the same.

(I’m avoiding feature promises that might not be in your region/plan. When in doubt, open “Help/Docs” inside your app.)

Teach It Your Style Once; Get Better Answers Forever.

You don’t get the output you deserve — you get the output you specify.
Teach your AI your rules first, then ask for work.

Here’s the purpose in one breath: the 25 prompts + a 120-word Style Card create a repeatable system so ChatGPT-5 mirrors your voice, priorities, and decision rules — consistently. Less rewriting. More results.

Be precise, or be ignored.

Your move (two minutes):

1. Paste Prompts #1–25 into a fresh chat.
2. Drop in 2–3 writing samples.
3. Generate and save your Style Card.
4. Use the Answer Shape [ (Too Long; Didn’t Read) TL;DR → Why → Steps → Pitfalls].
5. Run the Self-Review and Revision. Publish.

Teach the process once. Reuse it forever.

If this helped, bookmark the prompts, share your Style Card wins, and keep iterating. You’re not just prompting — you’re installing a thinking style your AI can execute on demand.

What if ChatGPT could become your personal mentor? Discover how in Build a Personal Mentor AI: The Ultimate System Prompt Framework for Growth is your next must-read. Unlock the AI that works for you.

Thank you for reading!