This is another example from Google’s Gemini 3 API docs. I feel like the average user doesn’t read through that kind of documentation so you’ll miss these nuggets of prompting wisdom. And even if you do read through them, maybe you need the reminder (I know that I usually do…)
The idea behind these examples is that you can leverage Gemini 3’s advanced thinking capabilities to improve its response quality for complex tasks by prompting it to plan or self-critique before providing its final response.The following example is for explicit planning. Add this to the end of your prompt(s)
Before providing the final answer, please:
-
-
- Parse the stated goal into distinct sub-tasks
- Check if the input information is complete
- Create a structured outline to achieve the goal
-
The following example is for self-critique. Add this to the end of your prompt(s)
Before returning your final response, review your generated output against the user’s original constraints.
-
-
- Did I answer the user’s intent, not just their literal words?
- Is the tone authentic to the requested persona?
-
You likely won’t need these for simpler tasks but they can definitely be helpful as things get more complex.
Still Here? You’re Probably Serious About This.
Get new AI insights each week—short, useful, and just uncomfortable enough to make you better.
Comments