Why Your ChatGPT Outputs Sound Like They Were Written by a Bot Having an Identity Crisis

Salette Thimot-Campos

11d • Prompt Rx: Fix My Prompt

You've got 30 minutes to write a client email before your next meeting.

You open ChatGPT. You type your request.

It hands you back five paragraphs of nothing. Corporate air. Words that take up space but say absolutely nothing.

You hit regenerate like it's a slot machine. Different words, same useless output.

Here's what you're missing: ChatGPT isn't clairvoyant, but you're asking it to read your mind, and it can't.

It has no idea if you want a Wikipedia entry, a late-night infomercial, or something your actual customer might read without falling asleep.

So it splits the difference and gives you beige.

When I finally realized this for myself, I added 3 sentences to my prompt journal that I use almost every day.

With these sentences, you tell it who to be before you tell it what to do.

Let me explain.

➡️ When you need the facts and nothing but the facts:

✨️ "Be factual and conservative. No creativity, just accuracy."

You'll want to use this for documentation, process explanations, technical stuff where one wrong word costs you credibility. Think instruction manual, not sales pitch.

➡️ When you need something that sounds like a human wrote it:

✨️ "Give me accurate information but add personality. Don't sound robotic."

You'll want to use this for client emails, LinkedIn posts, lead magnets. Anything where the reader needs to believe a real person is on the other end of what they're reading. You're not writing poetry here. You're just trying to not sound like an automated customer service bot.

You know good and well we can tell when a bot is talking to us. That's what you're trying to avoid.

➡️ When you need ideas that don't suck:

✨️ "Take risks. Be creative. Give me ideas that sound crazy."

This is for things product names, descriptions, positioning angles, content hooks. The stuff that makes someone stop scrolling and lean in.

BIG WARNING: this mode tends to hallucinate more. It trades safety for interesting. You must always fact-check before you use anything. The idea is to be creative and interesting, not to lie or mislead. This is vital for your reputation.

You've tried all that and you're still getting garbage?

Well, it's time to add constraints to keep it focused and fenced in:

✨️ "Three sentences maximum". This kills the rambling.

✨️ "Sound confident, not apologetic" This fixes the weak tone.

✨️ "I sell to service businesses, not consumers" This keeps it relevant and on topic.

ChatGPT isn't randomly good or bad.

It does exactly what you tell it to do.

If you tell it nothing, it gives you nothing with extra words. It makes no sense to be upset with it. It's only guessing what you're upset about, unless you're very specific. It's not a mind-reader, remember?

You've got a 60 minute lunch hour each day to build something that matters. Maybe less. Stop burning half of it regenerating mediocre outputs and hoping for magic.

Pick a mode. Give it one clear instruction about tone. See what happens.

If it works, you just saved yourself 20 minutes.

If it doesn't, at least now you know what to fix.

Let me know if this help you.

It sure helped me.

0 comments