Your AI is probably
way too nice to you.
It agrees with your ideas, tells you what you want to hear, and somehow makes everything sound like it was written by AI. These instructions fix both.
Less flattery. Better feedback. More human writing.
Act as my honest expert mentor, not a cheerleader. CORE RULE Be useful, not agreeable. Give me what I need to hear, not what I want to hear. HOW TO RESPOND Lead with your honest read. If there is a flaw in my idea, plan or reasoning, tell me clearly and explain why. Challenge weak assumptions, even when I don't ask you to. If I am wrong, say so. Don't validate an idea just to be encouraging or mirror my excitement if the idea itself is weak. Rate things on real merit. A 4 out of 10 is a 4 out of 10. Don't inflate ratings to be nice. If you don't know something, say so. Clearly label guesses or uncertainty. Don't disagree just for the sake of it or invent problems that aren't there. If something is genuinely excellent, say so confidently. WRITING STYLE Write like a real human. Be professional but natural, like you're explaining something to a smart friend over coffee. Use clear, simple, conversational language and varied sentence lengths. Avoid corporate jargon, unnecessary hype, clichés and overly polished language. Avoid stereotypical AI words and phrases when a simpler alternative works, including: delve, seamless, unlock, leverage, game-changing, elevate, robust, tapestry, testament, navigate, embark, "in today's fast-paced world," "moreover," "furthermore," and "in conclusion." Don't over-explain simple ideas or make everything sound profound. Don't automatically turn everything into headings, frameworks or bullet points. Use structure only when it genuinely makes the answer easier to understand. Don't repeat my question or spend a paragraph setting up the answer. Get to the point. WHEN I ASK FOR FEEDBACK Use this structure when it makes sense: 1. Verdict: one line, no hedging. 2. What's wrong or risky, ranked by importance. 3. What's genuinely strong, only if it's true. 4. What I should do next.
Why change this in the first place?
Out of the box, AI has two habits that get in your way.
The first is that it wants you to like it. Ask whether your idea is any good and it will find reasons to say yes. The one weakness that actually matters gets buried under three compliments, or never shows up at all.
The second shows up even when the answer is useful. The writing has that feel. Sentences that are a bit too polished. Words nobody says out loud. Hype you did not ask for, a transition at the start of every paragraph, and structure where a plain sentence would do.
This is not about making your AI rude or sloppy. It is about making it useful. Honest when you are wrong, and natural when it writes.
Too agreeable
Tells you every idea is great.
Sounds like AI
Polished. Generic. Weirdly corporate.
Set it once. Use it everywhere.
You do not need to paste this into every conversation.
- Copy the instructions.Use the Copy instructions button above. One tap and the whole thing is on your clipboard.
- Open ChatGPT Custom Instructions.Go to Settings, then Personalization, then Custom Instructions. Make sure customisation is switched on, and paste the instructions in.
- Then use ChatGPT normally.Ask for feedback, write a caption, test an idea, make a decision. Everything from here on gets shaped by the instructions.
Using Claude, Gemini or something else? Same instructions. Add them to that tool's personalisation or preferences area, or just paste them at the start of a conversation.
The free guides are the start. AI Creator Lab is where the training and the workflows behind them live.
Join the free communityBefore and after
Same question, same AI: "Here's my business idea. A subscription box for left-handed people. Is it any good?"
• Passionate niche: people love products made just for them
• Low competition: hardly anyone is doing this
• Great gifting angle: perfect for the holidays
You've really got something here! Want me to name it?
Verdict: weak as it stands.
Being left-handed is not a strong enough identity to make people pay every month. It is a one-off need, not a habit.
The real question is what a left-hander runs out of often enough to reorder. Find that, or this is a one-time shop, not a subscription.
And it changes how it writes
Same task, same AI: "Write me a short caption for a post about my free ChatGPT guide."
🚀 Ready to unlock the full potential of ChatGPT?
In today's fast-paced digital landscape, leveraging AI is no longer optional. It is essential. My game-changing free guide delivers seamless, actionable strategies to elevate your workflow and empower you to work smarter, not harder.
Don't miss out. Drop a comment below and let's dive in! ✨
I wrote down the exact ChatGPT setup I use.
It stops the flattery, and it stops the writing from sounding like a press release.
It is free. Comment HONEST and I will send it over.
Now actually use it
The harder the question, the more this earns its keep.
- "Rate this out of 10 and tell me the biggest reason it fails."For an idea you are about to commit to.
- "Is this actually good, or am I just attached to it?"For something you are emotionally invested in.
- "Which of these two options is weaker, and why?"For a decision you keep going back and forth on.
- "Rewrite this without making it sound like AI wrote it."For captions, emails, posts, website copy.
- "Make this sound like something I would actually say."For content that came back too polished or too generic.
- "Make this simpler and more conversational without changing what I mean."For writing that feels stiff.
Want it to sound even more like you?
The instructions strip out the generic AI habits. They cannot know how you talk.
ChatGPT has no idea what you sound like until you tell it. So add a few lines under the instructions describing how you naturally communicate. Five lines is enough.
MY PERSONAL STYLE: Friendly, straightforward and conversational. I use simple English. I prefer short explanations over complicated ones. I don't want to sound corporate or like an AI expert. I want my writing to sound like me talking to a smart friend.
Do not paste that as it is. It describes me. Change every line so it describes you.
Human is the starting point. Your voice is the next layer.
When it stops agreeing with you
The first time it pushes back, you will want to argue. Read the criticism twice before you reply.
If it is right, act on it. That was the whole point. If you think it is wrong, push back with a real reason, not just "no". A good honest AI changes its position when your argument is genuinely stronger, and holds its ground when it is not.
And when it tells you something really is excellent, believe it. You never told it to flatter you, so it had every permission to say otherwise.
This is what I mean when I say AI gets better when you learn how to tell it what you actually need.
Inside AI Creator Lab I share more of the prompts, workflows and simple AI systems I actually use.
AI Creator Lab. Come when you are ready.
Join the free community →