Want to improve your ChatGPT results? Try these 6 easy mindset shifts

ChatGPT is helping people do all sorts of things, from planning holidays and writing emails to debugging code, navigating mental health wobbles, and crafting dating app bios. Whether you think that’s brilliant or deeply concerning isn’t the point here (though I definitely share some of your reservations about its life advice). The fact is, it’s a tool used by millions, and that number is growing fast.

What’s surprising is how few people really understand how it works. I don't say that to sound superior, it makes sense. Tech companies aren’t exactly shouting about the downsides. Nobody tells you upfront that it can make stuff up or that, underneath it all, it’s essentially a very fancy prediction machine. But these things are pretty important to know.

The good news is that you don’t need to spend money on a course or learn advanced prompt engineering to get better results. I mean, you can. But if you don’t have the time, money, or bandwidth, a few small mindset shifts might be all you need. These are the ones we’ve found most useful after months of hands-on testing, research, and helping others get to grips with AI tools, as well as understand their limitations.

Mindset shift 1: Assume it’s making everything up

This isn’t being cynical. It’s understanding how the system works. AI tools like ChatGPT are prone to what’s called hallucination, which essentially means making things up because they’re generating text based on prediction, not truth.

I’d recommend reading up on why this happens, but the key thing to remember is: it does happen. A lot. That doesn’t mean you shouldn’t use ChatGPT; it just means you need to bring your critical thinking to every response. Especially if something sounds too perfect, too slick, or too confident, because it's probably too good to be true.

The way to navigate this without feeling frustrated is to always treat its responses as first drafts, not final answers. Always double-check names, stats, and facts, even if you’re using an AI tool that searches the web in real-time.

Mindset shift 2: Use it like a different perspective machine

We all have biases, and unfortunately, ChatGPT is excellent at confirming them. Ask a leading question, and a lot of the time it’ll happily serve up a similar viewpoint.

But it doesn't have to be that way. With the right approach, it can also help you challenge your thinking. Ask it to play devil’s advocate. Request a counterargument. Invite it to channel a philosopher, critic, or expert with a different worldview from you.

So rather than asking something like “What’s the best way to think about...?” Try: “What are three completely different ways to approach this?”

You’ll probably get more interesting results. You might learn something new. But most importantly, you’ll nudge it out of its default setting, which still seems to be agreeable, generic, and a bit too eager to please you.

Mindset shift 3: It’s a mirror, not a mind-reader

There’s a lot of fuss online about how to engineer the best prompts to feed your favorite AI tool. You'll find advice packed with frameworks and secret formulas. And you can get stuck into that if you like. But the key thing you need to know is simply that the quality of the output really does depend on the clarity of your input.

That means it pays to be as specific as possible about what you need. Think of it like briefing a very fast, slightly overconfident intern who's desperate to help you out. You need to be crystal clear from the beginning about your goal, audience, tone, and format. Otherwise, it’ll fill in the blanks with its own assumption,s and they’re often not what you're looking for.

For example, this would be incredibly vague: "Write a blog post about AI in education."

Whereas this would get you much better results: “Write a 500-word blog post about the future of AI in education, aimed at curious but non-technical readers in a friendly and informative tone. I’ll send you some examples of past work next so you can see how I write.”

You can, and probably should, say even more. We've got a whole guide about how to get ChatGPT to write like you. But that’s a good starting point.

And if it doesn’t nail it the first time, don’t panic. Treat it like a conversation. Give feedback, refine your ask, and keep going until you get what you're looking for. Or enough of what you're looking for to then work on the output independently, so you can make it sound like your own.

Mindset shift 4: You’re the director

ChatGPT can take on a lot of different roles. If you want it to think and write like a social media copywriter or a therapist or a film critic, you can literally tell it to “act as…” and that'll shape your results accordingly.

So, although you are just prompting in a different way, it might help to see yourself as directing a performance. And the more detailed your direction, the better the outcome.

Think of it like casting and briefing a character in a scene. Give it tone, contex,t and constraints. You’ll still need to guide it – never assume it truly understands the knowledge or nuance of that role – but it can be surprisingly good at getting the basics right.

Mindset shift 5: It’s a tool, not a creative replacement

I know, I know. People keep saying, “It’s just a tool.” In fact, I've heard it so many times it's starting to feel a little meaningless.

But the point is that yes, ChatGPT can do all sorts. But it doesn’t have your taste, your judgment, or your lived experience. That’s why the best results come when you use it as a partner or a guide, or a brainstorming buddy. To generate ideas, structure your thinking, or tidy up a draft, rather than doing everything for you.

It’s brilliant at helping you over the first hurdle, too, especially if you're staring down a blank page. But the creative decisions, like defining the style, the tone, getting to the bottom of what you're really trying to say, and refining it all at the end, all of those important decisions are still yours to make.

Mindset shift 6: Avoid the emotional trap

ChatGPT can feel weirdly wise, warm, and even comforting. But it’s really important to remember it's not your friend, therapist, or mentor. It’s a text prediction engine. That’s it.

It’s easy to forget that, especially when you’re stressed or stuck or lonely. But assigning too much humanness to it can lead you to over-trust its answers or become a little too reliant on its feedback. That’s where things get blurry.

You might also like

• Sam Altman doesn’t think you should be worried about ChatGPT’s energy usage - reveals exactly how much power each prompt uses
• Google can now generate your AI videos more quickly than ever
• Sam Altman thinks superintelligence is within our grasp and makes 3 bold predictions for the future of AI and robotics: ‘We are past the event horizon’