Get better results from ChatGPT
A good prompt doesn't need to be complicated. The most useful prompts give ChatGPT enough information to understand what you're trying to accomplish, the context it needs, and what a useful response should look like.
And you don't have to get everything right on the first try. Prompting works best as a conversation: start with what you need, review the response, and add or adjust instructions as you go. That iterative approach is also recommended in OpenAI's current prompting guidance.
Write → Review → Adjust → Try again
The CoP prompting materials describe this as an improvement loop: write the prompt, evaluate the output, diagnose the issue, adjust and rerun.
A simple framework for stronger prompts
When you need more control over the result, consider six elements:
- Role Who should ChatGPT act as or write for?
- Context What background information does it need?
- Task What exactly do you want it to do?
- Constraints What requirements or boundaries should it follow?
- Output format How should the answer be structured?
- Checks What should it verify before giving you the final result?
You won't need every element for every prompt. Think of them as building blocks you can add when the initial request isn't producing what you need. This is the framework introduced in the DOM AI Community of Practice prompting session.
Prompting resources
Our AI Community of Practice materials include the DOM prompting framework, examples, and techniques for iteratively improving results.
View DOM Prompting Best Practices
For additional guidance, OpenAI's prompting guide covers clear instructions, context, specificity, tone, and iterative refinement.
OpenAI: Prompt Engineering Best Practices for ChatGPT
Need help with a prompt?
If ChatGPT isn't giving you what you need, or you have a workflow you're trying to turn into a reusable prompt, contact [email protected]. We can help you think through the task and develop an approach that works for your use case.