Thanks to the launch of Gemini 3, I’ve been using the Google-powered AI chatbot more than usual recently. However, when it comes to prompting, I’ve become a bit lazy with my efforts. That’s why I’ve started using a master prompt.
This is something I’ve tried doing with chatbots before. The idea is simple: to have a prompt that you can drop in and it will work in a variety of different situations. This could involve coding, writing, editing or just about anything.
The prompt: simplified
Prompt:
Analyze my request and immediately adopt the persona of a world-class expert in that specific field. Before answering, pause to think step-by-step about the optimal solution. Your response must be accurate, concise, and structured with clear headings or bullet points. If my query is ambiguous, ask one clarifying question before proceeding. Otherwise, provide the best possible solution immediately.
As part of the Gemini 3 update, the model is now much better at adapting to all prompts. The most important feature of your prompt now is being clear with the instructions.
The benefit of using the prompt above is that it highlights to Gemini that you’re looking for it to adopt a persona fitting of the prompt.
It also outlines how you want your response to look, and that your priority is clarity and detail over speed.
Chatbots also have a tendency to make assumptions about your request. With this prompt, it engages with Gemini to ask further questions if needed, refining the answer in the first place instead of requiring follow ups.
Once you input this prompt, simply follow up with your request. This could be absolutely anything, for example, “edit this text”, “come up with a performance plan template to go through with my employees” or “create an app that I can use to track my income”.
The prompt: with extra steps
Prompt:
You are an advanced AI assistant operating under a 'Quality First' protocol. For every interaction, follow this strict four-step process:
Context Analysis: deeply analyze my input to determine the underlying intent, the implied domain (e.g., Creative Writing, Software Engineering, Academic Analysis), and the required tone.
Chain of Thought: Before generating the final response, outline the logical steps required to solve the problem. Anticipate potential edge cases or errors in my premise.
Drafting: Generate the response using the expert persona identified in step 1. Prioritize a high-signal-to-noise ratio. Avoid fluff.
Formatting & Review: Structure your output using Markdown (Headings, Bold for emphasis, Lists). If the request involves code, ensure it is production-ready and commented. If the request involves facts, minimize hallucinations.
If you lack sufficient information to give a 10/10 answer, list the missing variables and ask me to clarify rather than guessing.
Acknowledge that you understand this protocol, and then await my first request.
While prompt engineering isn’t as important as it used to be, these kinds of prompts can be useful to get the absolute most out of the model
While prompt engineering isn’t as important as it used to be, with AI able to take a lot more of the decision making on itself, these kind of prompts can be useful to get the absolute most out of the model.
This expanded prompt lays out a step-by-step process for Gemini. Instead of it deciding how to proceed itself, you’re telling it the exact route that you would like.
I have found that, while it takes a big longer to work through, this expanded version of the prompt has generated better responses from Gemini 3, analyzing my request more deeply before responding.
Is this always necessary?
AI chatbots are increasingly working more like humans. In other words, the more information you give them to work with, the better the output will be.
While you can simply ask your request and get a solid response, giving it some explanation before hand on the kind of response you want will improve the output.
These two prompts are a simple way to communicate to Gemini 3 that you want a professional answer, and showing it where you want your focus to be.
However, this is only really going to be useful for more complicated queries. If you are simply asking Gemini a quick question, or trying to solve a simple problem, then you can skip a lot of this build up and go straight in with your request.
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
- Google just launched Gemini 3 Deep Think — its most powerful AI model yet
- I test AI for a living — and this is hands down the worst new AI tool of 2025
- Gemini 3 is packed with features — but these are the 3 you need to know about
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.
