Best AI Prompts for Market Research (2026)

Prompts for market research#

Act as a market research analyst. Build a TAM, SAM and SOM estimate for [DESCRIBE YOUR PRODUCT AND TARGET CUSTOMER] using only the source excerpts I paste below. For every input, quote the source line, its publisher and its date. Show the arithmetic step by step, list each assumption separately, and mark any input I did not supply as MISSING instead of estimating it. [PASTE YOUR SOURCE EXCERPTS]

Act as a competitive intelligence analyst. From the competitor pages, pricing pages and filings I paste below, build a feature matrix for [LIST YOUR COMPETITORS] with one row per capability and one column per company. Put "not stated" where the sources are silent, link each cell to the quote that supports it, and then list the gaps and the customer types that no competitor addresses, labelled as your inference. [PASTE YOUR COMPETITOR SOURCES]

Act as a survey researcher. Design a research plan for learning [DESCRIBE THE DECISION YOU NEED TO INFORM] from [DESCRIBE YOUR TARGET RESPONDENTS]: the hypotheses to test, screener questions, a questionnaire free of leading wording, the sample size and how respondents will be recruited, and how you would tell a real pattern from noise in the answers.

Act as an industry analyst. Using only the dated reports and data tables I paste below, summarise how [DESCRIBE YOUR INDUSTRY] changed over [TIME PERIOD]. Separate what the sources state, what they forecast and what you infer. Quote any forecast with its publisher, date and stated assumptions, and flag where two sources disagree. [PASTE YOUR REPORTS AND DATA]

Act as an investment research associate. From the company filings and funding announcements I paste below, tabulate funding rounds, dates, amounts and named investors for [LIST YOUR COMPANIES OR SECTOR]. Do not calculate valuation multiples unless both the valuation and the revenue figure appear in the sources, and flag every figure that comes from a press report rather than a filing. [PASTE YOUR FILINGS AND ANNOUNCEMENTS]

What to add to any market research prompt#

• The source material itself. Anthropic advises placing long documents at the top of the prompt, above the query, and Google advises supplying context first, with instructions after it.
• A quote-first step. Anthropic's guide says to ask the model to quote the relevant parts of the documents before carrying out the task. It makes each figure traceable to a line you can check.
• A closed world. Google's guide suggests treating the provided context as the limit of truth. Tell the model to answer "not in the sources" instead of filling a gap from memory.
• Dates and vintages. Census data are tied to a specific vintage (a reference year), so state the year each source covers and ask the model to flag mismatched periods.
• The output shape. Name the table columns, the units and the currency. Google's guide says you can specify the format of the response, such as a table.
• Search controls, if your tool browses. Anthropic's web search tool supports allowed and blocked domain lists and returns citations, so restrict it to the publishers you trust.

A worked example#

The first prompt, filled in for a made-up product. The excerpts are placeholders for data you retrieve yourself.

Act as a market research analyst. Build a TAM, SAM and SOM estimate for a scheduling app for independent dental practices in the United States using only the source excerpts I paste below. For every input, quote the source line, its publisher and its date. Show the arithmetic step by step, list each assumption separately, and mark any input I did not supply as MISSING instead of estimating it.

Excerpt 1 (US Census Bureau, County Business Patterns, retrieved 2026-10-22): [PASTE THE ESTABLISHMENT COUNT FOR OFFICES OF DENTISTS, WITH ITS REFERENCE YEAR]
Excerpt 2 (competitor pricing page, retrieved 2026-10-22): [PASTE THE PUBLISHED PER-LOCATION PRICE]
Excerpt 3 (my own survey of 40 practices, run in September 2026): [PASTE THE SHARE THAT USE PAPER OR SPREADSHEETS]

A good result must contain:

• Each number carried with its quote, publisher, date and reference year, so you can open the source and match it.
• Arithmetic you can redo by hand, with TAM, SAM and SOM defined in one line each and tied to a stated filter.
• Assumptions kept apart from sourced inputs, such as the share you could realistically win, labelled as assumptions.
• "MISSING" wherever you did not paste an input, such as a growth rate.
• No figure that appears in none of your excerpts.

If competitor facts will appear in marketing, the FTC says comparative advertising is legal as long as it is truthful and that advertisers must have proof behind express and implied claims. Keep the dated source for each claim, and have counsel review comparisons before they run.

The full prompt library is on our prompt library page.

Limits of this guide#

The vendor guides explain how to supply sources and ask for quotes; none of them tests market research specifically, and we did not test these prompts. Nothing here tells you whether a TAM method is sound for your market. That needs a judgement about definitions and filters that the model cannot make for you. This page is about desk research from supplied sources; for launch positioning and channel plans, see our marketing plan prompts. We could not confirm exact request limits for the EDGAR or BLS APIs from the pages we read.