How to prompt ChatGPT: 10 tips for better answers

I've complained on more than one occasion that I wish my partner knew what I wanted without me having to tell him. But I've finally accepted a universal truth: nobody's a mind reader. Not even an AI chatbot trained on the entire internet.

When you prompt ChatGPT, you have to say the quiet part out loud. Sure, you can tell it to write your Substack newsletter and get something. But it probably won't be filled with the wry sensibility your audience knows you for. ChatGPT needs explicit instructions and context to generate the kinds of outputs you can actually use. 

I've been using ChatGPT since the early days. Based on my experiences, and the experiences of the entire Zapier team, I've rounded up the top 10 tips to help you prompt ChatGPT to get better answers. The same tips also work in Claude, Gemini, Copilot, and most other chatbots.

Table of contents:

• How to prompt ChatGPT

• Automate ChatGPT with Zapier

• How to prompt ChatGPT: FAQ

How to prompt ChatGPT

Here are 10 tried-and-true tips to help you prompt ChatGPT for better results. Not every tip will apply to every prompt. The first few cover the fundamentals—context, examples, format—and are worth internalizing regardless of what you're using ChatGPT for. The latter ones are more situational: model selection, chaining, and a few phrases that reliably shift how the AI responds. Start where it's useful, and go from there. 

1. Use a simple ChatGPT prompt skeleton

If you're new to prompt engineering, use this prompt skeleton as a starting point. It covers the basics of what ChatGPT needs to know to generate useful outputs. 

• Define the role. Tell ChatGPT who it should be. For example, an editor, analyst, or customer support lead. 

• Explain the task. Detail what you want ChatGPT to generate. For example, you can tell it to brainstorm viral social media campaigns to promote your new business. 

• Give it context. Share helpful details, including background information, constraints, or examples to use as inspiration. Instead of dumping your every thought into the chat, be careful to give ChatGPT only relevant information (more on this in a bit). 

• Define the output. Specify what the output should look like, including the format and length. If you're asking ChatGPT to write something, be sure to give it guidelines on how to write like you. 

Your prompt doesn't need to be overly detailed or complex. Here's an example: 

2. Give ChatGPT context

Just like humans, AI does better with context. Think about exactly what you want the AI to generate, and provide a prompt that's tailored specifically to that. Here's an example of what I mean. 

A word of caution: There's a fine line between giving ChatGPT just enough context and too much context. Research has found that AI models don't weigh all information evenly, and performance "grows increasingly unreliable as input length grows." Large chunks of your directives, especially the middle bits, may even get lost in the noise. So be conservative with what you feed ChatGPT. It'll take some tweaking to find that happy middle ground, and that's ok. 

3. Show ChatGPT examples (good and bad)

Providing examples in the GPT prompt can help the AI understand the type of response you're looking for (and gives it even more context).

For example, if you want ChatGPT to reply to a user's question in a chat-based format, include a previous example of a conversation between the user and the agent. You'll want to end your prompt with "Agent:" to indicate where you want the AI to start typing. You can do so by using something like this:

Examples can also be helpful for math, coding, parsing, and anything else where the specifics matter a lot. If you want to use ChatGPT to format a piece of data for you, it'll be especially important to give it an example. Like this:

You can also tell ChatGPT what to avoid by showing it negative examples or referencing previous results it generated for you that you didn't like. I find this particularly helpful if ChatGPT's responses start drifting. 

4. Tell ChatGPT the length of the desired response

When crafting your ChatGPT prompts, it's helpful to provide a word count for the response so you don't get a 500-word answer when you're looking for a sentence (or vice versa). You might even use a range of acceptable lengths.

For example, if you want a 500-word response, you could provide a prompt like "Write a 500-750-word summary of this article." This gives the AI the flexibility to generate a response that's within the specified range. You can also use less precise terms like "short" or "long." 

For code or data, name the format. For example: JSON only, markdown table, or CSV then chart.

5. Set up ChatGPT once (so you prompt less)

If you find yourself repeating a lot of the same instructions, ChatGPT offers a few features that allow you to set it (your baseline instructions) and forget it. Here's a high-level overview of the ones worth setting up. For detailed instructions, check out Zapier's in-depth guide for how to use ChatGPT. 

• Custom instructions: This lets you program ChatGPT with instructions you want it to always keep in mind when generating a response. If you use ChatGPT for copywriting, for example, this is a good place to tell it your role, how formal or informal you want replies, what to avoid, and any other guidelines to be sure it writes like you. 

• Memory: ChatGPT automatically picks up on details and preferences across chats to tailor its responses. But you may want to remove irrelevant memories so they don't interfere with what you want ChatGPT to actually remember. 

• Projects: For recurring work—for example, weekly recaps or customer support macros—I use ChatGPT Projects to keep related chats, memories, custom instructions, and other reference material in neatly contained groups. This way, you can jump between projects and have ChatGPT simultaneously context switch with you.

• Custom GPTs. If you're on a paid tier, you can use natural language to create a custom ChatGPT (also known as a GPT)—each programmed with its own set of instructions—or use prebuilt ones. If you're on a free plan, you can use GPTs, but you can't create them. 

6. Pick the right model type for the job

ChatGPT offers a handful of models designed to accelerate at different types of tasks. The model names and categories are constantly evolving. There are over a dozen working models, each with a different ideal use case (some models have been superseded by the latest OpenAI model, but they're still available in the API). And that's just ChatGPT. Other AI chatbots might run on another AI model or a combo of models. 

Here's what I recommend: no matter which chatbot you're using, match the prompt to the model. For example, if you want ChatGPT to run a multi-step analysis of last quarter's earnings, compare AI vendors, or debug a process, use GPT-5.5. It's currently OpenAI's best model for advanced reasoning and logic. If you want ChatGPT to do in-depth research on treatments used to support dogs with hip dysplasia (and you have the budget), use GPT-5.5 Pro. 

7. Ask ChatGPT to interview you first

Sometimes you don't know what you don't know. ChatGPT can help flesh these details out. Open the chat with something like this:

Optionally, you can then ask ChatGPT to summarize your answers and use that to write a fresh prompt in a new chat. Anecdotally, I've found this helps with any messy context window issues.

8. Chain prompts for big projects

For complex tasks—like debugging code, planning a launch, or anything with several moving parts—you usually get better results from a chain of prompts than from one mega-message. Each step has one job, and you review before you move on. For example, instead of asking ChatGPT if your Python code is thread-safe, add specific instructions to generate logic. Like this: 

9. Use phrases that steer ChatGPT's reasoning 

Sometimes it's just about finding the exact phrase that OpenAI will respond to. Here are a few phrases that folks have found work well with OpenAI to achieve certain outcomes. 

• "Let's think step by step." This makes the AI think logically and can be specifically helpful with math problems.

• "Show your work, then give the final answer." This can help if the AI keeps arriving at inaccurate conclusions.

• "In the style of [famous person]." This is helpful for matching styles.

• "As a [insert profession/role]." This helps frame the bot's knowledge, so it knows what it knows—and what it doesn't. 

• "Explain this topic for [insert specific age group]." Defining your audience and their level of understanding of a certain topic will help the bot respond in a way that's suitable for the target audience.

• "For the [insert company/brand publication]." This helps ChatGPT understand which company you're writing or generating a response for, and can help it adjust its voice and tone accordingly.

10. Have ChatGPT write the prompt

Sometimes you know the outcome you need but not the exact wording ChatGPT responds to. When that happens, describe the job in plain language and ask ChatGPT to draft the reusable prompt for you. Here's an example:

Copy what ChatGPT returns and run it in a new chat.

Automate ChatGPT with Zapier

If you live in ChatGPT (or any other chatbot), and want the model to act—not just respond—use Zapier to connect it with the rest of your tech stack. With Zapier MCP, you can take action across 9,000+ apps directly from your chatbot. It can update a Salesforce record, post to Slack, add a row to Google Sheets, or kick off a multi-step workflow across your entire app stack, right from the chat thread.

Zapier handles every OAuth connection behind the scenes, so your credentials never end up in the conversation. You control which apps your AI can touch, and you can cut access from one place if anything changes.

How to prompt ChatGPT: FAQ

I've been writing about ChatGPT for a while now, and a few questions come up over and over. Here are answers to the ones I get asked the most.

How do you write a good prompt for ChatGPT?

This varies based on your task, but, at a minimum, be sure to include these in your ChatGPT prompt: Give ChatGPT a role, a clear task and context (audience, constraints, and pasted text), and any output rules. For complex work, I recommend using a reasoning model and asking ChatGPT for step-by-step thinking before outputting the final answer.

Do these prompting tips work in Claude or Gemini?

Yes. ChatGPT is the focus here because it's my go-to chatbot, but context, examples, format, and iteration matter across all AI chatbots.

Related reading:

• AI prompt templates to use with your AI chatbots

• How to write AI art prompts

• Claude vs. ChatGPT: Which is best? 

This article was originally published in January 2023, with contributions from Reid Robinson and Elena Alston. The most recent update was published in May 2026.