ChatGPT has a hidden set of creativity switches — here’s how to use them

ChatGPT is an enthusiastic mimic. You just have to give it the right persona to emulate. One way to break out of the beige tone of a standard reply is to deliberately inject a stylistic voice into the prompt. You can do this by referencing a writer, a character, a cultural tone, or even a fictional vibe. It’s less about accuracy and more about mood-setting.

Let’s say you want a simple to-do list to help get through a rough Monday. Normally, ChatGPT will return something like: “1. Eat a healthy breakfast. 2. Review your schedule. 3. Tackle high-priority tasks.” Not bad.

But now say, “Make me a Monday to-do list in the style of a drill sergeant who’s had too much coffee.” Suddenly it’s barking, “Rise and shine, maggot! That inbox won’t clear itself!” The tasks are still useful, but the tone shifts from bland to bracing. Or try, “Make it sound like a Victorian governess addressing a chaotic household,” and you’ll get: “Kindly restore order to the parlour before Aunt Edith arrives.”

Sign up for breaking news, reviews, opinion, top tech deals, and more.

I’ve used this switch to amuse myself with everything from turning grocery lists into Shakespearean dialogues to getting the weather from a 1970s sci-fi narrator to making my calendar into a series of riddles. What you’re doing is giving ChatGPT a creative costume to wear. The key is to get specific. The weirder the tone, the looser ChatGPT gets.

This switch relies on asking ChatGPT to be aware of its own tendencies and deliberately override them. You can nudge it to break the default format by directly calling out what it usually does and asking it not to.

For example, you might say: “Don’t write this like a typical AI summary. Write it like you’re a snarky ghost with no memory of being alive but strong opinions about pizza.” Or: “Most of your answers are calm and neutral. I want a hot take that sounds like it would start a family argument.” The trick here is to acknowledge what the model typically gives you and frame the request as a deviation.

You may like

I tested this with a friend who wanted to write a funny Instagram caption for her dog. Her first attempt at a caption for her dog sitting in a laundry basket was “Just helping with the laundry! #DogLife.” Not memorable.

So she asked ChatGPT to “Write an Instagram caption from a dog who’s convinced the laundry basket is a spaceship and has deep beef with the vacuum.” ChatGPT’s next attempt: “Commander Bark reporting from the Sock Nebula. Mission: avoid the Roomba at all costs.” That’s a post people actually remember.

ChatGPT always tries to help directly. It does not naturally consider alternative vantage points unless you ask for them. Changing the narrator or point of view in the prompt opens up new ways for the AI to consider answers, whether it’s through the eyes of an inanimate object, a pet, or simply from a long way away. ChatGPT responds by reorganizing information around the new narrator rather than rewriting the same answer with different adjectives. The effect is often surprising because it reveals angles you would not have considered on your own.

For example, imagine you are deciding whether to buy a new pair of running shoes. If you ask directly, the model will give you competent, but not very inspiring, buying advice. If you flip the perspective and tell the model to answer as if it were your future self who has already run hundreds of miles in the shoes, the advice becomes more narrative and reflective. It might describe how the new shoes changed your morning routine, or how they helped you survive a muddy trail run and made you feel lighter on days when you were tired.

That kind of advice can turn a simple shopping question into a small personal story that captures the feeling behind your decision. The model’s baseline tone is practical, but a perspective shift convinces it to behave like a storyteller instead of a rulebook.

Once you start to see ChatGPT not as a single personality but as a flexible system that responds to mood cues, you realize how much room there is to personalize it. Most people don’t explore that space. They assume that because ChatGPT’s output sounds smart, it must know the best version of what you want. But often it’s just giving you the most average version of what everyone else has asked for.

These creativity switches let you reclaim the weirdness. They turn ChatGPT into more of a scene partner than a fact-delivery tube. You can still ask for recipes, emails, summaries, and errands. But when you flip the right switch, the model starts to surprise you. It becomes a little more surprising in its answers, and that might make it more useful, if a little more chaotic, than its baseline approach.