For months, I used ChatGPT almost daily as a friendly partner. Yet I somehow ignored one tiny setting that could have made everything smoother: Custom Instructions. Once I finally started experimenting with it, my answers became sharper, more relevant, and more “me.” It was like giving ChatGPT split personalities as I could dial into different voices whenever I wanted. Even if you are a ChatGPT beginner learning to write prompts, set up this underused feature and shape ChatGPT into a personal assistant that actually sounds like it knows you.
What are ChatGPT personas?
A simple setting to change how ChatGPT thinks and talks
If you’ve ever wished ChatGPT understood your writing style, tone, or work preferences, you need to think about personas. They’re essentially profiles that define how ChatGPT behaves and responds. In OpenAI’s terminology, this is done through Custom instructions, a feature tucked under Settings > Personalization.
There’s one simple field that fulfills two functions:
- What would you like ChatGPT to know about you to provide better responses?
- How would you like ChatGPT to respond?
By filling this field thoughtfully, you turn ChatGPT from a generic assistant into a more predictable collaborator. It remembers your preferences for every new chat. Very much like an alter ego who knows your favorite tone, formatting style, and level of detail.
You can choose any of the provided ones. But you will get more out of ChatGPT with your own personal instruction. When I first tried it for a newsletter project, I set mine up as:
You are a newsletter writer who writes clear, practical articles for general readers at an eighth-grade reading level and uses short paragraphs.
Notice how I made it more specific (“eighth-grade reading level”) instead of generic (“clear, practical articles”). You can also experiment with the ChatGPT personality setting, but I keep mine as Default.
Suddenly, ChatGPT’s suggestions sounded much closer to my voice. Instead of overexplaining, it started producing structured, readable paragraphs that felt ready to drop into a draft.
Why setting a persona makes a big difference
Make AI stop guessing and start aligning with your goals
Before I customized ChatGPT, I found myself repeating the same instructions every session: “Write in a friendly, professional tone. Keep it concise. Use American English.” It was repetitive.
With Custom instructions, those details become baked-in context. Every new chat begins with the right assumptions, freeing me to focus on the ideas instead of the setup. The difference in output was immediate.
For example, when I brainstormed a headline idea, ChatGPT stopped suggesting clickbait like “10 Secrets to Boost Productivity Fast” and offered more relevant ones like “How to Use AI Tools to Reclaim Your Creative Focus.”
The feature works because it makes ChatGPT work the same way as colleagues do. They usually perform better after working with you for a few weeks. ChatGPT remembers everything about you, so it soon stops overexplaining and starts anticipating what fits you best.
How to create your own ChatGPT persona
Start with your real-world role, then layer on tone and detail
Creating a persona takes just a few minutes. Open Settings > Personalization > Custom Instructions, then fill in the field. The trick is to be specific without overloading ChatGPT. The text field has a 1500-character limit.
Here’s an idea for a simple formula that works for me:
- Start with who you are: your role, experience level, and what kind of help you want.
- Then, layer in how you want your responses i.e., your tone, level of detail, and formatting rules.
For example, a writer might say:
You are a freelance tech journalist who writes tutorials and explainers for everyday readers. You prefer concise, actionable paragraphs and plain English. Follow NYT title case for headings.
A teacher might say:
You are a patient educator explaining complex topics in simple language. Use relatable examples and short summaries after each explanation.
Once saved, these instructions automatically apply to all new chats until you edit or replace them.
Managing multiple personas
One setting at a time, but there are smart workarounds
So far, ChatGPT only lets you have one active persona in your Custom instructions at a time. That works fine if you mostly use ChatGPT for one purpose, but not if you juggle roles like many of us do. I am a writer by morning and a creative brainstormer by afternoon.
I found three easy ways around that limitation:
- Custom GPTs: If you’re a Plus or Team user, create multiple saved versions of ChatGPT under Explore GPTs > Create a GPT. Each can have its own persona—like “Newsletter Writer,” “Creative Partner,” or “Teacher.”
- Persona Library: Keep your favorite personas in a Notion page or Google Doc. Copy and paste the one you need before starting a chat.
-
Prompt-based switching: You can switch personas temporarily with each individual by using richer ChatGPT prompt techniques:
Act as my Editor persona and review this draft for clarity.
It won’t overwrite your main settings but will behave in that role for the session.
For serious multitaskers, Custom GPTs are the easiest long-term fix. Now, I am crafting a Devil’s Advocate Custom GPT to manage my assumptions.
Tips for fine-tuning your persona
The secret is specificity and iteration
Creating a great persona isn’t a one-and-done task. Some results are far from impressive. So, tweak it over time as you notice how ChatGPT interprets your requests.
Be specific: Instead of “Write clearly,” say “Write at an eighth-grade reading level and use short paragraphs.”
Add what you don’t want: For instance, “Avoid emojis” or something like “Respond to this question only if you find information that can be attributed to a citable research paper.”
Stack context: You can layer roles dynamically. For instance, ask ChatGPT to “act as a devil’s advocate,” without altering the base persona.
Revisit your persona every few weeks as your projects or tone change. Remember, ChatGPT and all AI bots do make mistakes.
Personalization is the fastest way to make ChatGPT truly yours
I still obsess over finding the perfect prompt formula. But like others, my problems usually come down to context. Without knowing who I am or what I want, ChatGPT’s output can swing wildly from verbose to robotic. Custom instructions solve this to an extent. At the very least, every chat starts with the right assumptions.