Dec 25,2025 
You want to know what your customers think before you build something for them. That makes sense. Talking to real users takes time, and sometimes you need quick feedback on an idea before committing resources. Synthetic users offer a way to simulate those conversations, letting you test assumptions and explore different perspectives without scheduling a single interview.
ChatGPT can help you build these simulated users. With the right prompts, you can create personas that respond to your questions, react to your product concepts, and surface objections you might not have considered. The process is straightforward once you know how to structure your requests.
This guide walks you through building synthetic users in ChatGPT, from basic persona creation to more advanced setups using custom projects. We will also share where this approach works well and where it falls short.
What a Synthetic User Actually Does
A synthetic user is a simulated person that an AI model generates based on the characteristics you define. You tell ChatGPT who this person is, what they care about, what frustrates them, and how they make decisions. Then you ask questions as if you were interviewing them.
The model draws on patterns from its training data to generate responses that fit the profile you created. You get something that resembles a conversation with a user who matches your target audience.
People use synthetic users for product feedback, messaging tests, onboarding flows, and early-stage research. The goal is rapid iteration, not a replacement for real user research.
Setting Up a Basic Synthetic User
Start with a prompt that gives ChatGPT enough context to maintain a consistent persona. You need to define the person and then tell the model how to behave.
Here is a template that works well:
“You are Sarah, a 34-year-old marketing manager at a mid-sized software company. You have two children under 5 and find yourself constantly juggling work and home responsibilities. You are skeptical of new tools that promise to save time because you have been burned before. You prefer simple solutions over feature-rich platforms. When I ask you questions, respond as Sarah would. Stay in character and answer from her perspective.”
After you send this, ChatGPT will respond as Sarah. You can then ask questions like:
- “What is the hardest part of your workday?”
- “If I told you this app could cut your meeting prep time in half, what would your first reaction be?”
- “What would make you trust a new productivity tool?”
The responses will follow the profile you created. You can push back, ask follow-up questions, and probe for deeper reactions.
Building Stronger Personas with More Detail
Basic prompts work, but richer context produces better results. Consider adding these elements to your persona setup:
- Background and history. Where did they grow up? What was their career path? What past experiences shape their current preferences?
- Current situation. What does a typical day look like? What pressures do they face? Who do they answer to at work?
- Goals and frustrations. What are they trying to accomplish? What keeps getting in the way?
- Decision-making style. Do they research extensively or buy on impulse? Do they trust reviews, ask friends, or rely on gut feeling?
- Attitude toward your product category. Have they tried similar products? What did they like or hate?
The more specific you get, the more grounded the responses become.
Using ChatGPT Projects for Persistent Synthetic Users
If you want to run multiple sessions with the same synthetic user, ChatGPT’s Projects feature helps. Projects let you create workspaces with custom instructions and uploaded documents that persist across conversations.
You can upload background materials about your product, your target market, or even sample customer feedback. ChatGPT will use this information as context when generating responses. The custom instructions field lets you define your synthetic user once, and every new chat in that project will maintain the persona.
Projects support up to 200,000 tokens of context, which gives you room to include product documentation, competitor analysis, or user research summaries. This creates a richer foundation for your synthetic user to draw from.
Prompts That Surface Useful Feedback
The quality of your synthetic user depends on what you ask them. Vague questions get vague answers. Here are prompts that tend to produce actionable insights:
- “Walk me through the last time you tried to solve this problem. What did you do first?”
- “I am going to describe a feature. Tell me your honest reaction, including any doubts.”
- “What would stop you from recommending this to a colleague?”
- “If this cost $15 per month, would that change your interest? Why or why not?”
- “What would I need to show you in the first 30 seconds to keep your attention?”
Push for specifics. Ask “why” when you get surface-level answers. Treat the conversation like a real user interview.
Where ChatGPT Synthetic Users Fall Short
ChatGPT generates responses based on statistical patterns in text. It does not carry memory between sessions unless you use Projects, and even then, the persona is reconstructed from your instructions each time. The model does not have lived history or genuine preferences. It simulates what someone like your persona might say, drawing on general patterns rather than specific behavioral data.
This works fine for brainstorming and early exploration. It struggles when you need accuracy or when small nuances in user behavior matter.
When You Need Human Accuracy: Evelance
For research that requires reliability beyond conversation simulation, Evelance offers something different. Evelance predicted real human responses with 89.78% accuracy in testing. In one case, 7 Evelance personas matched 23 real people in under 10 minutes.
The difference comes down to how the models are built. Standard synthetic users generate responses based on category averages and random attribute combinations. Evelance personas have internalized behavioral data until it became part of how they respond. Each persona carries identity, memory, and situational context. They measure 12 psychological dimensions for each persona and draw from over 2 million personas.
The result is a persona that knows who they are before your test begins. They hold opinions, carry history, and arrive with context already in place. This matters when you need to predict how real people will actually respond, not how a general category of people might respond.
Bringing It Together
ChatGPT gives you a fast, accessible way to create synthetic users for early-stage research and product thinking. Define your persona with enough detail, ask pointed questions, and use Projects when you need persistence across sessions.
When your decisions depend on predictive accuracy rather than general exploration, tools like Evelance offer a foundation built on real behavioral data. The right choice depends on what you need the synthetic user to do for you.