This content originally appeared on DEV Community and was authored by Anjana Jayaweera
ChatGPT 5 was released on August 7, 2025 and like many others, I wanted to figure out the best way to talk to it. Social media is full of tips, and I wanted to write this guide in a way that everyone can test it out for themselves, rather than following anything blindly.
The Starting Point
Think of it like meeting someone new. You don’t just start asking random questions. You introduce yourself, share a bit about your interests, and set the tone for the conversation.
Why can’t we apply the same principle to ChatGPT as well?
Step 1: Introduce Yourself to ChatGPT
OpenAI has an option to do this via “Custom Instructions”. You can find this either under “Settings → Personalization → Custom Instructions”, or by clicking the “Customize ChatGPT” link at the bottom-left of your profile image.
Here are two demo videos from OpenAI showing how it works
The first key section is:
“What traits should ChatGPT have?”
Here, you decide how ChatGPT should reply. You can choose from suggested tags or write your own.
Next is:
“Anything else ChatGPT should know about you?”
If someone wants to know you better, what would you tell them about you?
You can write anything here. There’s no right or wrong way of writing this section but just be clear, concise, and honest. I personally keep it simple. I have organised mine into sections as follows:
- Professional Background
- Mindset
- Communication Style
- Formatting Preferences
- Level of Detail
- Hobbies
- Current Location
- Current Focus
Example :
I have a strong technology and product background, with 15+ years in software engineering and leadership.
I think and communicate best through examples, analogies, and step-by-step explanations.
I value clear structure, so I prefer bullet points, numbered lists, or other organised formatting.
I enjoy creative problem-solving, entrepreneurship, and exploring ways to build meaningful projects from simple ideas.
I like practical, forward-thinking advice that’s direct and actionable, not sugar-coated.
I balance a busy professional life with personal pursuits like focusing on my family, cricket, photography, and living a healthy life, so time efficiency is important.
Tip:
Keep your “current focus” updated. ChatGPT adapts its responses based on what matters to you now.
Optional:
Delete saved memories if you want a fresh start, but if you find them helpful, keep them.
Step 2: Ask ChatGPT for a Clear Prompt Structure
Now that you have introduced yourself, it’s time to get to know about ChatGPT. Start a new chat and paste the following prompt:
“Hey Chat, I want to understand how to communicate better with you. Is there any recommended way I should give instructions? Can you give me a template to follow?”
Once you do this, based on your custom instructions, ChatGPT will give you a template catered specially to you. It usually looks like this:
Task: [What you want the AI to do]
Context: [Background, constraints, goals]
References: [Links, examples, docs]
Format: [Bullet points, table, JSON, step-by-step]
Persona: [Choose or customize]
Evaluation Criteria: [Tone, length, technical depth]
Extra Instructions: [Optional style notes or things to avoid]
There’s no one-size-fits-all. You can write prompts in natural language or bullet points. The key is clarity. Personally, I like bullet points, but sometimes natural language feels easier.
Example :
Option 1 — Natural language:
Act as a CTO. I want a 90-day technical roadmap for launching a SaaS MVP for small accounting firms.
Context: Bootstrapped budget, need rapid validation, using a remote dev team.
Constraints: Focus on speed, low cost, and maintainability. Bullet points only.
Output: Table with “Phase”, “Key Actions”, “Tools/Tech”, “Expected Outcome”.
Option 2 — Bullet style:
Task: Create a 90-day technical roadmap for launching a SaaS MVP for a small accounting firm.
Context: Bootstrapped budget, need rapid market validation, using a remote development team.
Constraints: prioritise speed, low cost, and maintainability.
References: SaaS MVP best practices, lean startup methodology, and case studies of bootstrapped SaaS launches.
Format: Table with columns: Phase, Key Actions, Tools/Tech, Expected Outcome.
Persona: CTO — provide blunt, strategic advice focused on delivering a functional MVP quickly while balancing technical debt with future scalability.
Evaluation Criteria: Roadmap must have actionable steps, be achievable in 90 days, and clearly identify quick wins. Must account for technical trade-offs and cost efficiency.
Extra Instructions: Use bullet points within the table cells. Avoid over-engineering; focus on features essential for launch and validation.
Use whichever style feels comfortable to you.
And that got me thinking, where else could this format be applied?
Scaling this structure across various Use Cases
Initially, I assumed it was mostly for those “Act as…” style prompts. Turns out, I was wrong. Once I started experimenting, I realised the structure works across a wide range of use cases, including:
Brainstorming and Ideation — Generating fresh concepts, creative angles, or product ideas.
Technical Planning and Architecture — Designing systems, workflows, and high-level blueprints.
Decision Support — Weighing pros and cons, evaluating trade-offs, and recommending actions.
Content Creation — Writing articles, scripts, marketing copy, or creative pieces.
Training and Documentation — Producing clear instructions, guides, and learning material.
Risk Assessment and Compliance — Identifying potential pitfalls and regulatory considerations.
Optimisation and Review — Refining ideas, improving workflows, and stress-testing strategies.
Example : Decision Support
Act as a Cloud Solutions Architect. Compare three cloud hosting options for running a high-traffic headless CMS for a travel website.
Context: We expect 5M monthly visitors, mostly from Asian countries, and require sub-200ms TTFB with strong CDN coverage in the region. The solution must handle high traffic, be reliable, and scale cost-effectively over three years.
References: AWS, GCP, and Azure hosting documentation, pricing, and regional performance benchmarks.
Format: A comparison table with columns: Provider, Pros, Cons, Estimated Monthly Cost, Scalability Rating.
Evaluation Criteria: Identify the best choice for balancing cost, performance, and scalability over a 3-year horizon. Must consider global and regional latency, CDN presence, and infrastructure reliability.
Extra Instructions: Include relevant performance benchmarks if available; ignore small-scale hosting providers or niche solutions. Ensure recommendations account for traffic spikes during seasonal peaks.
This is where the real power of this format clicked for me. It’s not just about making ChatGPT “roleplay” a persona, it’s about giving it a structured lens to tackle almost any problem.
Try It Yourself!
Give these steps a shot and see how it works for you. AI is evolving every second, and I’m all ears to learn from your experiments and insights too!
Leave a comment below with your experience. What worked, what didn’t, and how you overcome the flows.
What this experiment led to
I spotted an opportunity. Why not turn this process into a lightweight app that can generate prompts on demand?
So, I vibe coded a “ChatGPT Prompt Builder App” using Cursor AI. Here’s how it looks.
Here’s the Link
https://chatgpt-prompt-builder.vercel.app
This content originally appeared on DEV Community and was authored by Anjana Jayaweera