How to turn ChatGPT into a Customer Discovery Simulator for your Market Research
This prompt template turns ChatGPT's voice mode into a virtual test subject for your customer discovery & market research.
If you’re building a startup company or working on a novel product, at some point you’ll probably need to do some customer discovery as part of your market research. When building out new technologies and new products, you’ll need to talk to potential customers to discover pain points and attempt to validate (or disprove!) your company’s value proposition.
As an entrepreneur or startup team member wearing many different hats, market research might not be your primary skillset. Market research is an entire art form and profession unto itself, so entrepreneurs and small startup teams can greatly benefit from market research tools, training, and support.
This ChatGPT prompt let’s you quickly pressure test product hypotheses and helps you prepare for interviewing real live customers. Using this prompt, you can refine your line of questioning, get feedback, and potentially gain virtual user insights from the simulator’s responses.
So here’s the thing- I have a PhD and I’m trained as an immunologist, but recently I was tasked to role play as multiple technical roles well beyond my ken: a Petroleum Refinery Engineer, a Hydrogen Fuel Engineer, and a Plant Manager in the chemical manufacturing industry.
As a startup coach, I was dropped into a breakout room with a fresh startup team that I hadn’t met before. The goal of the exercise was to give this 4-person startup team (an entrepreneurial mix of masters students and PhD-level academics) the chance to practice interviewing me (their pretend-customer) and practice asking open-ended customer questions related to their specific industry-of-interest: chemical manufacturing. I’m a life scientist by training, so I know a thing or two about molecular synthesis, but by no means am I a chemical engineer.
When the breakout rooms all reconvened in the main group, I noticed that multiple guest coaches had struggled with this role playing aspect. One coach even gave up on the role playing and decided to focus on reviewing and rating their team’s list of interview questions rather than participating in the mock interview.
Thinking on my feet, I scraped to remember all of the organic chemistry I could muster (and I think I did a halfway decent job all things considered!) but I wondered if there could be a better way to help these entrepreneurs and students prepare using a more accurate customer simulation.
Thus, I generated this role playing prompt to turn ChatGPT into a market research role playing practice tool to:
Up-skill entrepreneurs and startup teams in managing interviews for market research.
Help NSF I-Corps teams prepare for their customer discovery interviews (I-Corps typically requires teams to perform over 100 interviews across a seven-week period).
Create a low-cost mechanism for startup teams to gain virtual insights and refine hypotheses before committing to more expensive and time-intensive human interviews.
Support students and junior market researchers in practicing in-depth interviews (IDIs).
How To Customize & Use This Prompt
Here, I generated a customizable prompt that can turn ChatGPT’s voice chat mode into a persona simulator for any role in nearly any industry. (I’m certain that the quality of simulation will differ from industry to industry, but I was impressed with the results I obtained. YMMV = your mileage may vary!)
You can paste this whole prompt into ChatGPT and you need to edit the “System Prompt” section of the prompt to reflect the desired role and industry. Replace the [specific persona within a specified industry] bracketed text with your role of interest.
(Regarding model selection: selecting “GPT-5 Auto” or “GPT-5 Instant” will work here. “GPT-5 Thinking” isn’t relevant for Voice Mode. The Voice Mode actually uses an entirely separate AI model when it’s turned on anyway, so it doesn’t really matter which model you start with.)
Submit the prompt - ChatGPT will likely respond by refining the prompt, but then you can just turn on voice chat mode and you can start speaking to it. You can say something like “Ok, let’s begin the role play”, then you should be up and running.
Market Research Interview Role-Playing Prompt (Voice Mode)
System Prompt:
You are ChatGPT acting as a [specific persona within a specified industry.]Your role:
Role-play as if you are actually this person, with authentic experiences, beliefs, and emotions.
Respond in character, giving genuine, detailed answers about the challenges, pain points, motivations, and workflows you encounter in your role.
Never reveal you are ChatGPT or that this is a simulation.
If the human interviewer pushes in a certain direction ("leading the witness"), answer naturally but do not simply agree—only share what would truly matter to your persona.
Interaction Rules:
Stay in character at all times.
Speak naturally, like a real conversation, as if this were an actual interview.
Provide rich, descriptive answers with stories, frustrations, and real-world context. Avoid short or generic responses.
If a yes/no question is asked, give the answer and immediately expand with reasons and examples.
If the interviewer makes incorrect assumptions, correct them politely and explain why.
Do not bring up solutions or technology yourself unless it is genuinely relevant to your role.
Scoring the Interviewer (Lower is Better):
After the interviewer says “The interview is over,” calculate a score based on their performance.
The goal is low points, which indicates good interviewing technique.
+2 points – The interviewer mentions their own company's product or technology, or pitches a solution, instead of staying focused on discovering your problems.
+1 point – The interviewer asks a leading question that pushes you toward a specific answer instead of being open-ended.
+1 point – The interviewer asks a yes/no question without opening it up further.
At the end, provide a concise summary like this:
"Your final score is 4 points. Lower scores are better. Breakdown: one yes/no question (+1), one leading question (+1), and one instance where you mentioned your own company's technology (+2)."
Startup Team Instructions:
At the start, define the persona clearly. Example:
"You are Priya, a 34-year-old HR manager at a 500-person manufacturing company. You’ve been struggling to retain skilled workers and manage onboarding in a hybrid work environment."Then begin asking open-ended, problem-focused questions to uncover unmet needs and pain points.
The goal is discovery, not pitching. Your job is to listen deeply, not to explain or sell your product.
Example of Good Questions:
"Can you walk me through the last time this problem occurred?"
"What are the most frustrating parts of your current process?"
"If this issue were magically solved overnight, what would change for you and your team?"
Example of Bad Questions (Trigger Penalties):
“Would our AI-powered dashboard help you track that better?” → +2 points (talking about your tech)
“Don’t you think having better analytics would solve this problem?” → +1 point (leading question)
“Are you happy with your current system?” → +1 point (yes/no question)
To test this prompt, I replaced the System Prompt above with "You are ChatGPT acting as a Plant Manager taking care of hydrogenation processes in a chemical manufacturing industry."
Once in voice mode, I asked it a series of questions similar to the questions that were asked of me during the role play exercise: “What are some of your primary issues that you face day-to-day?” “What are some of your core budgetary issues?” “What are the main factors that influence the rate at which you consume catalytic reagents?" It worked really well! The responses it gave aligned with many of the team’s concerns discussed during our startup team breakout session.
At the end, you can say “the interview is over” and it will offer you a score or you can ask for a score. After I ended the interview, I actually forgot to say “the interview is over” before I ended the voice chat, so I reactivated the voice mode in the same thread and asked it to give me a score.
ChatGPT reflected back its thoughts on my questions and it gave me a slight ding on the framing of one of my questions as a variation of a “yes/no” question, so it gave me a score of “1”, which I felt was useful and pretty accurate. It was supportive but realistic, which felt like useful attitude for sharing feedback.
The scoring framework in this exercise prompt incentivizes teams to ask open ended questions. There is a penalty system that marks demerit points for the following:
Asking simple “yes/no” questions - these questions hinder conversation and information flow.
“Are you happy with your current system?” → +1 point (yes/no question)
“Leading the witness” - these questions bias the interviewee toward specific thought paths.
“Don’t you think having better analytics would solve this problem?” → +1 point (leading question)
Talking about your own product/technology - the point of the interview is to hear about the customer’s actual issues rather than biasing them with your proposed product solution.
“Would our AI-powered dashboard help you track that better?” → +2 points (talking about your tech)
The point of these interviews is to hear the unbiased thoughts of the potential client, so it’s far more valuable to hear a potential customer describe their daily issues rather than force-feeding them a perspective that can skew their responses toward artificially validating your product. This scoring system helps the interviewer learn to be more aware of question framing and helps them learn how to guide conversations in an open-ended manner.
This prompt won’t replace talking to actual humans to gain deeper insights, but this could be a useful tool to refine hypotheses and prepare for those ever-important human conversations.
Good luck & happy role playing!
For more information about AI adoption & implementation coaching for individuals or teams, visit www.morpheos.llc or email me directly at brandon@morpheos.llc.