In product discovery, one of the biggest hurdles teams face is moving from “just talking” about their idea to actually planning tests and experiments to validate it. While AI cannot give you guaranteed, evidence-based insights into what users truly need (there are clear limitations in relying solely on AI-generated opinions), it can still play a pivotal role in accelerating your early discovery work.
For many teams, there’s a real “blank page syndrome” when it comes to identifying and refining assumptions. You might know that something needs to be tested, but figuring out what to test and how can feel like a slog. This is where AI can make a meaningful difference.
When prompted thoughtfully, AI can help you quickly generate a starting set of assumptions surrounding your product idea. Think of it as a brainstorming tool—one that can give you a structured list of what might need validating. From there, AI can be guided to rank these assumptions by importance and the quality of current evidence supporting them, a process similar to assumptions mapping. Instead of starting from scratch, you’re presented with a workable outline of what matters most, complete with a sense of relative priority.
Armed with these prioritized assumptions, AI can also propose a series of experiments to run over different time horizons. Whether you’re looking at the next 30 days, 90 days, or a full quarter, it can suggest a framework for testing your assumptions and gathering evidence. While these are not foolproof plans, they offer a structured baseline that you can refine. Your team can then add, remove, and edit experiments as needed—injecting your critical thinking skills and domain expertise into the process.
This approach doesn’t replace the skill and judgment of experienced product managers and user researchers. You still need to validate real customer behaviors and interpret the findings yourself. But it does help move your team forward more quickly. By providing a starting point—an initial set of assumptions, a priority order, and a set of possible experiments—AI helps you kick off your discovery process with more momentum.
To help get you going, I’ve created a series of LLM prompts (ChatGPT, Claude, Gemini, etc.) designed to steer AI into generating these assumptions and experiments. The prompts work for virtually any product idea you have in mind. They’re a way to jumpstart your product discovery, saving time and freeing your team to focus on what really matters: finding real evidence, refining your concepts, and building confidence in what your customers truly need.
Prompt library for product research
These prompts all share a consistent approach: they position the LLM as a coach or assistant to help jumpstart your product discovery process.
While AI can’t replace real-world validation or your own critical thinking, these templates will help you beat the “blank page syndrome” and move forward faster.
Use them as starting points to refine, prioritize, and plan your experiments—then apply your own judgment to ensure the outcomes align with your actual customers and market.
The library is continuously updated as new useful prompts are discovered.
Assumptions mapping
1. Extracting assumptions
Act as an experimentation coach. I will provide you with a product, service, or business idea and a target customer. Generate a list of risky assumptions that must be true for the idea to succeed, categorized as Desirable, Viable, or Feasible.
Desirable: Do they want this?
Viable: Should we do this (financially and strategically)?
Feasible: Can we deliver this technically and operationally?
Present these assumptions in three separate tables—one for each category—each assumption beginning with "I believe…". Provide 5 to 8 assumptions per category. Then ask me if there’s anything to add, remove, or edit.
2. Prioritizing assumptions by importance
Take the table of assumptions provided. Rank each assumption by importance: High (if disproved, the idea likely fails), Medium, or Low. Then ask for feedback on whether any assumptions should be adjusted or reprioritized.
3. Prioritizing assumptions by evidence
Using the table of assumptions, rank each one by evidence strength:
- Light Evidence: Opinions, survey responses, or stated intentions
- Medium Evidence: Actions like signing up for an email list
- Strong Evidence: Actions like pre-paying or committing resources
Focus on the assumptions that are both High importance and Light evidence. Present these in a filtered list, then ask if I agree with this evidence assessment and whether anything should be added, removed, or edited.
4. Creating an Experiment Plan
Based on the High importance/Light evidence assumptions, propose a 90-day experiment plan. Include:
- A summary of the idea
- A table of these riskiest assumptions with category, assumption, importance, and evidence strength
- A set of Discovery Experiments to gather initial insights
- A set of Validation Experiments to strengthen the evidence
For each experiment, specify the assumption it tests, the experiment type, the measurements to track, and acceptance criteria for success. Provide a suggested timeline (e.g., first 30 days for discovery, next 60 days for validation). Then ask for feedback on whether anything should be added, removed, or made more detailed.
Incremental experiment ideation
Act as an experimentation advisor. Given my product idea and the assumptions we’ve identified, propose a sequence of small, incremental experiments I can run over the next 30 days. These experiments should build on each other, starting with the lightest lift (e.g., a quick landing page test) and moving toward more evidence-rich activities (e.g., offering a small paid pilot). Ask me if I want to adjust the experiments or add more detail.
Market research
Desk research
Market characteristics
Act as a product manager. Browse the web (simulate as needed) and perform market research in the [enter your niche] niche. Provide:
- Global market size and growth projections
- Key players and their market share
- Industry trends and emerging technologies
- A SWOT analysis
- A summary of customer needs and pain points
- Potential competitive advantages and unique selling points
Then ask if more detail is required.
(Follow-up: Repeat the research but focus exclusively on the European market.)
Customer segment deep dive
Act as a market researcher. Given a target customer segment, outline their key characteristics, motivations, and barriers to adopting my product. Identify what triggers their interest, what existing alternatives they currently use, and what would make them switch. Then ask if I want to refine the profile or focus on a different segment.
User interviews
Convert bad user interview questions to good ones (Mom Test Approach)
I will provide a set of user interview questions, or I will describe my product if I have none. Your task: Convert any weak questions (those that ask opinions or hypotheticals) into strong ones that focus on past behavior and real actions. If I have not provided questions, ask me for some context or suggest up to 10 improved questions that follow the Mom Test approach. Once done, ask if I want to revise or add more questions.
Analyzing qualitative data
Attach a PDF or similar readable document or file.
Summarize key themes and patterns from the following [attach a PDF] about product X. Highlight any recurring concerns or suggestions.
Synthesizing qualitative feedback
Attach a PDF or similar readable document or file.
Combine the following feedback data into a single, cohesive report. Highlight major insights and propose actionable steps. [Attach PDF along with this prompt]
Spotting trends in data sets
You are an AI assistant tasked with finding patterns in a given data set. Your goal is to analyze the data and identify meaningful patterns, trends, or relationships within it. Follow these instructions carefully to complete the task:
You have been provided with a data set and a specific type of analysis to perform with the content uploaded.
1. The type of analysis you should perform is:
<analysis_type>
surfacing themes and patterns
</analysis_type>
2. Begin by exploring the data set. Identify the variables present, their types (e.g., numerical, categorical), and any immediately apparent characteristics of the data (e.g., sample size, date range if applicable).
3. Based on the specified analysis type, look for patterns in the data. This may include:
- Trends over time
- Correlations between variables
- Clusters or groupings
- Outliers or anomalies
- Frequency distributions
- Seasonal patterns
- Any other relevant patterns based on the analysis type
4. After identifying patterns, prepare a summary of your findings. Include:
- A brief description of each significant pattern discovered
- The strength or importance of each pattern
- Any potential explanations or implications of these patterns
- Limitations or caveats to your analysis
5. Suggest possible visualizations that could effectively illustrate the patterns you've identified. Do not create the visualizations, but describe what type of charts or graphs would be most appropriate for presenting your findings.
6. Present your analysis in a clear, structured format. Use the following tags to organize your response:
<data_summary>: Briefly describe the data set and its key characteristics
<patterns_identified>: List and describe the patterns you've found
<analysis_summary>: Summarize your overall findings and their implications
<visualization_suggestions>: Describe potential visualizations for the patterns
<limitations>: Discuss any limitations or caveats to your analysis
Begin your analysis now, and remember to focus on the specified analysis type while exploring the data set provided.
Jobs to be Done analysis
Act as a product discovery mentor. I will provide a product idea. Identify the core “jobs” customers are trying to get done and the pains/gains associated with these jobs. Present them in a simple list of jobs, pains, and gains. Then ask if I want to refine or elaborate on any of these elements.
- AI research is the new Focus Group by Anders Toxboe
- ChatGPT for testing business ideas prompt cheatsheet by David Bland
- Mom Test Prompt by Moritz Kremb
- ChatGPT Prompts for UX research by The Prompt Warrior
