Club Benefits
The latest AI models are faster, smarter and better at reasoning than ever. Regardless of which chatbot you prefer to use, if you ask it a question, you'll get an instant response. Ask it to draft an email, analyze ideas or break down complex problems and you'll get everything you need in seconds.
But there’s a catch: the quality of the answer often depends on how you frame the problem. I can't tell you how many times I've asked a chatbot a question without framing the prompt the right way, only to get back a frustrating and generic response.
But, the prompts I rely on most do something different — they tell the AI how to think. I call them "thinking prompts": a small library of framing techniques that change the model’s reasoning process, not just the output. Instead of asking a better question, you’re giving the AI a better mental model to follow.
These are the thinking prompts I reach for most — the ones that consistently produce better results.
1. The 'first principles' prompt
Prompt: "Explain this problem using first-principles thinking. Break it down to the most fundamental truths and rebuild the explanation from the ground up."
Article continues belowFirst-principles thinking pushes Claude to strip away assumptions, clichés and conventional wisdom. Instead of repeating the most common explanation, it has to break the problem down into its core components and rebuild the answer from the ground up. That often leads to responses that are clearer, more logical and more original.
This is especially useful when a topic feels overly complicated, full of jargon or shaped by “the way things are usually done.” By forcing the model back to fundamentals, you get an explanation that is easier to trust and easier to understand.
2. The 'contrarian' prompt
Prompt: "Challenge the common assumption behind this idea. What might skeptics or critics say?"
AI models are trained on large datasets that often reflect mainstream thinking. As a result, their default responses tend to reinforce common assumptions or widely accepted viewpoints.
This prompt interrupts that pattern by asking the model to step outside consensus thinking and actively search for weaknesses in an idea. By exploring what critics, skeptics or contrarians might say, the response becomes more balanced and intellectually rigorous. I use this one a lot when shaping my ideas or for arguments when blind spots might be hidden.
3. The 'expert panel' prompt
Prompt: Imagine a panel of experts discussing this problem. What would different experts disagree about?
You may have noticed that AI sounds confident, turning one idea into a simplified viewpoint. I don't find this helpful, which is why I created this prompt to encourage the model to simulate a conversation between different types of experts, each bringing their own priorities, assumptions and areas of expertise.
When multiple perspectives are introduced — such as a technologist, economist, psychologist or strategist — the answer becomes more layered. Instead of presenting one “correct” solution, the AI highlights where experts might disagree, where trade-offs exist and which factors matter most. You can tweak this one to the particular type of experts your needs require.
4. The 'simplify it' prompt
Prompt: "Explain this idea as clearly as possible for a beginner. Avoid jargon and use simple examples."
One of my favorite ways to use AI is to simplify things. But, you have to direct them to do that. Large language models often default to complex explanations filled with jargon, technical terms or unnecessary detail. That can make answers sound like they are coming from an expert, but much harder to understand.
This prompt forces the AI to simplify the concept and focus on the core idea. By removing the lingo or jargon and using relatable examples, the model has to translate complexity into something intuitive and easy to grasp. Another perk of using this prompt is that when the AI is pushed to explain something clearly for a beginner, it usually produces responses that are not only easier to understand but also more logically structured and useful for learning.
5. The 'improve the idea' prompt
Prompt: "Critique this idea and suggest ways it could be improved."
Although Claude is less of a people pleaser than other charbots, it will still default to being agreeable. When you present an idea, the model can help to reinforce it rather than evaluating it critically. This prompt shifts the AI into a more analytical role by asking it to identify weaknesses, blind spots and opportunities for improvement.
Instead of simply validating the concept, the model acts more like a constructive critic or editor. It examines the idea from multiple angles, highlights potential flaws and suggests practical ways to strengthen it.
6. The 'structured thinking' prompt
Prompt: "Analyze this problem step by step and explain your reasoning."
If you're tired of a chatbot jumping straight to the answer without slowing down to unpack how it got there, this prompt is for you. It slows the process down and encourages the model to work through the problem in a more deliberate, structured way.
By asking for a step-by-step analysis, you push the AI to break the issue into smaller parts, make its logic more transparent and build the answer in a sequence that’s easier to follow. That usually leads to responses that feel more organized, thoughtful and reliable. I use this one a lot for problem-solving in any situation where I need to understand the why and the reasoning behind the response. It's a really productive way to learn.
7. The 'real-world test' prompt
Prompt: "If this idea were applied in the real world, what challenges or trade-offs would appear?"
AI often produces answers that sound convincing but stay at the level of theory. This prompt forces the model to move from abstract ideas to really exploring real-world consequences.
By asking what would actually happen if the idea were implemented, the AI has to consider practical constraints such as cost, incentives, unintended side effects and trade-offs. That shift usually produces more grounded insights and reveals problems that might not appear in a purely theoretical explanation. I use this prompt a lot when I have a "great idea" that needs to be thought through from angles I may not have thought of.
Bottom line
Large language models don’t just respond to questions — they respond to how the problem is framed. This is true for even the best reasoning models. So when you guide the reasoning process upfront, the model has a clearer structure to follow. Instead of generating a quick answer based on patterns in its training data, it works through the problem more deliberately.
The result is responses that are often clear, more creative and less generic. Give these prompts a try and let me know what you think in the comments.
Follow Tom's Guide on Google News and add us as a preferred source to get our up-to-date news, analysis, and reviews in your feeds.
More from Tom's Guide
You must confirm your public display name before commenting
Please logout and then login again, you will then be prompted to enter your display name.
