12 prompts
AI prompts for market research
Updated
These AI prompts for market research turn raw inputs (reviews, survey responses, interview notes and competitor pages) into clear findings. Language models are good at sorting and summarizing text and weak at recalling current facts. So paste the evidence into the [BRACKETS] and ask the model to reason over it, not to recall market data from memory.
Customer research
Survey design
Use it before sending a customer or market survey.
You are a research designer. Goal: learn [RESEARCH_QUESTION] from [RESPONDENT_PROFILE]. Write a survey of 10 to 12 questions that takes under 5 minutes. Mix multiple choice, scales and 2 open questions. Avoid leading and double-barreled questions. For each question, note which decision the answer will inform. Remove any question that informs no decision.
Interview synthesis
Use it after 5 or more customer interviews.
Act as a UX researcher. Here are notes from [NUMBER] interviews with [SEGMENT]: [PASTE_NOTES_LABELED_BY_INTERVIEW]. Identify recurring themes. For each theme: a name, how many interviews mention it, 2 verbatim quotes with interview labels, and what it implies for our product or messaging. List contradictions between interviews separately.
Jobs to be done statements
Use it to frame research findings around what customers are trying to do.
You are a product strategist. Research inputs: [PASTE_FINDINGS]. Write 6 job statements in the form: When [situation], I want to [motivation], so I can [outcome]. For each, list the current workaround people use and the main frustration with it.
Open-ended survey coding
Use it when you have hundreds of free-text answers.
Act as a research analyst. Here are open responses to the question '[QUESTION]': [PASTE_RESPONSES]. Create a codebook of 6 to 10 codes with definitions. Assign each response one or more codes. Return a table of code, count and 2 example responses, then a 3-sentence summary.
Reviews and voice of customer
Review mining for a category
Use it to learn what buyers love and hate about existing tools.
You are a market analyst. Here are reviews of products in [CATEGORY] from public review sites: [PASTE_REVIEWS_WITH_PRODUCT_NAME_AND_RATING]. Tag each review with themes. Report the top 5 praise themes and top 5 complaint themes, with counts and verbatim quotes. Then list 3 unmet needs that appear across several products.
Support ticket themes
Use it to turn support data into product and marketing insight.
Act as a customer insights analyst. Ticket subjects and first messages from the last [PERIOD]: [PASTE_TICKETS]. Group them into themes, estimate share of volume per theme from the data given, and split each theme into: bug, confusion (fix with docs or UX), missing feature, or pricing question. Suggest 3 help articles or onboarding changes that would reduce volume.
Language bank for copywriting
Use it to collect the exact words customers use.
You are a copy researcher. Sources: [PASTE_REVIEWS_INTERVIEWS_SURVEY_ANSWERS]. Extract verbatim phrases customers use to describe: the problem, the moment of frustration, the desired outcome, and alternatives they tried. Return 4 lists of quotes with the source label. Do not paraphrase.
Competitors and market
Competitor positioning grid
Use it when you need a quick view of the landscape.
Act as a competitive intelligence analyst. Competitor homepage and pricing text: [PASTE_PER_COMPETITOR]. Build a table: competitor, target customer, main promise, pricing model and entry price, key differentiator, and an obvious gap. Then suggest the 2 axes that best separate the market for a positioning map, and where each competitor sits on them.
Market sizing logic
Use it to structure a bottom-up estimate you can defend.
You are a strategy consultant. I want to estimate the market for [PRODUCT] among [SEGMENT] in [REGION]. Build a bottom-up model as a list of assumptions: number of target accounts, share with the problem, willingness to pay, and price. Do not fill in numbers from memory. For each assumption, tell me which public source type could provide it (government statistics, industry association, public company filings) and what to search for.
Trend scan from sources I collect
Use it monthly to keep an eye on your space.
Act as a market researcher. Here are articles, posts and announcements from this month about [MARKET]: [PASTE_SNIPPETS_WITH_LINKS]. Summarize the 5 most important developments, why each matters for [MY_COMPANY], and one action to consider. Link each point to the snippet it came from.
Win and loss analysis
Use it after a batch of deals closes.
You are a revenue analyst. Notes from won deals: [PASTE]. Notes from lost deals: [PASTE]. Compare: deal size, buyer role, competitors involved, objections and decision reasons. Report the top 3 reasons we win and lose, with evidence, and 3 changes to sales or marketing.
Pricing research questions
Use it before changing prices, to learn what buyers will pay.
Act as a pricing researcher. Product: [PRODUCT]. Segment: [SEGMENT]. Current price: [PRICE]. Write a short survey using the Van Westendorp price sensitivity questions, adapted to our product, plus 3 questions on which features buyers value most. Explain how to read the results and the minimum number of responses I should collect before trusting them.
Run these automatically
When you add your site, Thig researches your competitors and writes an editable Competitor Analysis doc that every agent reads. The GEO agent also tracks how AI assistants describe your category. See the content gap skill to turn research into topics.
Frequently asked questions
Can ChatGPT do market research?
It can design surveys, analyze text you paste in (reviews, interviews, support tickets) and structure your reasoning. It should not be your source for market sizes or current facts, since it can be wrong or out of date.
What is the best prompt for competitor analysis?
Paste the competitor's own pages and public reviews, then ask for positioning, pricing model, strengths, weaknesses and gaps, with each point tied to the text you supplied.
How do I analyze customer reviews with AI?
Export reviews to text, paste them in batches and ask the model to tag themes, count mentions per theme and pull verbatim quotes. Spot check the counts yourself.