ChatGPT can be very useful, but its default approach to questions is often like a particularly dull textbook. That cautious approach to basic prompts makes sense as it is trying to appeal to as broad an array of people as possible, but it can mean you get frustrated with its answers if you want something beyond a sentence of generic paste.
But, with some creative thinking, I’ve found phrases you can insert into almost any prompt and get far more from the AI as a result. Here are some of the best prompt formats for getting sharper, more engaging responses from ChatGPT.
1. “Roleplay a situation where…”
ChatGPT’s dry, balanced, encyclopedia-esque tone is fine for facts, but if you want it to be more active in helping you, you can give it a job in the conversation. Suddenly, you have a method actor immersed in the role.
For instance, you might ask ChatGPT to: “Give me negotiation tips.” The list of strategies you get might help, but they don’t make you feel like you’re ready for that conversation.
Instead, you can tell ChatGPT to: “Roleplay a realistic situation where I’m negotiating a raise and you’re my boss who has some resistance to the idea, so I have to respond effectively.”
ChatGPT might start with: “As your manager, I appreciate your hard work, but budgets are tight this quarter. What makes you think now is the right time for a raise?” Now, you can actually practice what to say, and ChatGPT will continue the discussion with you from your boss’s perspective. It might not predict everything in the conversation, but you’ll likely be much better at answering those specific questions than otherwise.
2. “Given everything you know about me…”
ChatGPT is a machine-learning model, not a mind reader. But if you’ve been having ongoing conversations with it, it does remember the context of your chats. It has a memory it will add to automatically during your chats, and you can adjust it at will. Why not take advantage of that?
You could always ask, “What book should I read next?” and you’ll get some good recommendations, like 1984 or To Kill a Mockingbird. But they’re all from generic lists of bestsellers or classics, not tailored to you.
Check the ChatGPT memory or whatever you might have said about yourself in the current conversation and use “Given everything you know about me,” and the answer will be very different. You could even make it part of the prompt if you prefer.
So you might ask: “Given everything you know about me, such as my love of sci-fi, my preference for morally complex characters, and my need for a book that will emotionally torment me, what should I read next?”
Now, instead of the usual safe choices, ChatGPT might suggest: “Based on what you’ve described, you might love The Left Hand of Darkness by Ursula K. Le Guin. It’s got deep moral ambiguity, thought-provoking worldbuilding, and an emotional gut punch at the end.”
3. “What would an expert say about this?”
The sanded-down ChatGPT answers to questions make for bland reading at best. If you want depth or expertise, you have to ask for it. That means asking ChatGPT to “Tell me about the future of AI.” provokes a short list of the pros and cons of AI, with caveats and uncertainties at the forefront.
You’ll get much more intriguing responses by conjuring an expert on the matter and asking ChatGPT to “Tell me what an AI ethicist would say about the future of AI?” Of course, if you really want to spice up the answer you could invoke a couple of experts who might have different views by following up that prompt with: “Now contrast that with what a Silicon Valley venture capitalist would say.”
The conflict is much more engaging, even with ChatGPT’s otherwise cautious tone in responding with: “An AI ethicist might warn that unchecked AI development could lead to massive job displacement, privacy concerns, and algorithmic bias. Meanwhile, a Silicon Valley venture capitalist would argue that AI will supercharge innovation, create new industries, and increase global productivity.”
4. “Break this down step by step for a beginner.”
ChatGPT sometimes assumes you know more than you do, throwing jargon at you without context. Slowing it down makes a big difference. You can ask: “How does machine learning work?” But then, you’ll have to read a dense paragraph packed with technical terms. Unless you already understand them, the overall meaning might escape you.
For the uninitiated, a better way is to ask ChatGPT to “Explain how machine learning works step by step as if you’re teaching someone who has never coded before.” No more jargon in this answer: “Imagine you want to teach a dog tricks. You show it how to sit, and every time it does, you give it a treat. Over time, the dog learns that ‘sit’ means good things happen. Machine learning works in a similar way, except instead of dogs, it’s data, and instead of treats, it’s accuracy.” Machines as dogs is the kind of analogy far more people will get.
Easier to understand doesn’t have to mean simpler, either. Maybe you just need the right kind of person to explain it. Combine the simplifying prompt with the roleplaying idea, and you could get a much more entertaining answer from a specific persona that appeals to you. Try asking ChatGPT to “Explain machine learning like you’re a stand-up comedian performing for an audience that has no science background.” Or “Explain machine learning as if you’re a pirate telling your crew about a mysterious new kind of treasure.”
5. “Give me an unconventional take on…”
Ask a normal question, and you’ll get a normal answer. Ask for something unexpected, and things get interesting. You could ask ChatGPT to “Tell me about the impact of social media on society.” You’ll get a useful, if predictable, breakdown of the positive connections, technological advances, and entertainment possibilities, along with the rapid spread of misinformation, negative mental health impact, and authoritarian information control. It’s fine, but it’s nothing new.
Ask instead that ChatGPT “Give me an unconventional take on the impact of social media,” and you’ll learn that “While social media is often criticized for shortening attention spans, it may actually be improving micro-learning. Users are absorbing complex ideas in bite-sized formats, making knowledge more accessible than ever. Unfortunately, that includes maliciously incorrect information and corporate propaganda, too.”
