AI for market research: 7 tools, and prompts you can trust

In this article, I’ll walk you through the benefits of AI for market research, the best AI tools, the use cases we apply at Predictable Innovation and concrete prompts for customer, competitor, and market research.

Imagine, as a strategist, founder, innovator, or product marketer, if you could transform internet and interview data into actionable insights in record time without tedious manual processing.

Today, thanks to artificial intelligence (AI), this is possible.

AI-powered tools for market research can search and read large amounts of information on the web, perform data analysis, and deliver it through easy-to-digest formats thanks to natural language processing models, marking a new era for market research.

Start with what you need the research for. A client report, a go-to-market strategy, a first market and a go or no-go on an idea each need different steps, so the guide begins there. After that come the eight-step workflow, the seven tools, how to use ChatGPT and Claude, two mini guides on market sizing and competitor reviews, and prompts you can paste.

Let’s get started!

People use AI for market research for one of four goals: a client report, a go-to-market strategy and sales playbook, a first market to go after, or proof that buyers will pay before you build. Each ends in a different deliverable, so pick yours first: the steps differ. The best AI tool for market research depends on your goal, so each goal below shows the steps, the tool and the exact feature to use. Find your goal in the table, then jump to its steps.

Goal 1: “I need to deliver a market research report to a client.”

When you deliver a market research report to a client, the job is coverage and trust: every number traceable, every claim checked. Speed is the margin.

Prompts to use: market stats, sources to read first, patterns across interviews.

Goal 2: “I need a go-to-market strategy and a playbook my sales team can sell from.”

For a go-to-market strategy and a sales playbook, the job is to make four decisions with evidence: who to sell to, the enemy, the category and the main angle. Then turn them into the website, the pitch and the SDR kit. The steps follow the free GTM Strategy Playbook.

AI drafts; the team decides. The four decisions (steps 2, 3, 5, 6) stay with the people who will sell them.

Goal 3: “I need to choose which market to go after first.”

To choose which market to go after first, the job is a fair comparison: the same questions asked of every segment, with evidence, so the choice isn’t the loudest opinion.

Goal 4: “I want to know if anyone will pay for this before I build it.”

To know whether anyone will pay for an idea before you build it, the job is to look for proof against the idea, not for it. AI is good at agreeing with you, so ask it to find the reasons it will fail.

Warning for this goal: AI personas will tell you the idea is great. Use them to prepare questions, never as the answer.

You do market research with AI in 8 steps: frame the decision, do desk research, listen to buyers, talk to buyers, find the patterns, map the alternatives, decide, and build the deliverable. All four goals above follow the same 8 steps, and only the deliverable changes. The goals differ in which steps they lean on:

• Goal 1, delivering a market research report to a client, leans on steps 2 and 8.
• Goal 2, building a go-to-market strategy and a playbook your sales team can sell from, leans on steps 3 to 7.
• Goal 3, choosing which market to go after first, leans on steps 2 and 7.
• Goal 4, finding out if anyone will pay for it before you build it, leans on steps 3 and 4.

Steps 2 to 4 run at the same time, so you gather the evidence in parallel. Steps 5 and 6 each work from everything those three steps found. The only step AI never does alone is step 7, decide: your team makes the call.

The best AI tools for market research are ChatGPT, Claude, Gemini, Perplexity, Granola, Reddit Answers and Visualping: seven tools, one job each. The two big assistants, ChatGPT and Claude, now do the research themselves: ChatGPT deep research and Claude Research plan the searches, read the sources and return a cited report.

You can run the whole stack for about $40 to $60 a month: one $20 assistant plan, Gemini at $20, and free tiers for the rest. Prices are in US dollars per month unless noted, and they change often, so check the vendor’s page before you buy. Which tools you need depends on your goal. The chapter on what you need the research for shows the steps, the tool and the exact feature for each goal.

ChatGPT

ChatGPT is best for market research reports, data analysis and turning research into slides. Its deep research reads the web, your files and connected apps, then gives you a cited report you can export. You can start free, and Plus costs $20. ChatGPT can also help build your market ecosystem map by finding key players in your target industry or markets, a very helpful step for finding people to interview and building relationships that improve your go-to-market strategy. To see how to use it, read how to use ChatGPT for market research.

Claude

Claude is best for interview analysis, positioning work and repeatable research workflows. Its Research mode comes with citations, projects keep your context, and skills run the same method every time. In one project, Claude can build the report, the data tables and a summary deck as Word, Excel and PowerPoint files. You can start free, and Pro costs $20. To see how to use it, read how to use Claude for market research.

Gemini

Gemini is best as a second deep research opinion and for analysis across many interviews. Give Gemini Deep Research the same brief you gave ChatGPT: where the two reports disagree, that is where you check by hand. Gemini Notebook (formerly NotebookLM) takes up to 300 sources on Google AI Pro, including interview audio files, which it transcribes directly, and you can ask across all of them. It can also turn a set of interviews into an audio overview to listen to on the go. You can start free, and Google AI Pro costs $19.99.

Perplexity

Perplexity is best for quick, sourced answers: market stats, news and who is who. It browses the web as it answers, so you get up-to-the-minute information with the sources shown first. Pro search includes PitchBook, CB Insights and Statista data, and you can connect your own IBISWorld or Crunchbase account. Perplexity now leads with “Computer,” an agent for multi-step tasks. You can start free; Pro costs about $20 and Max costs $200.

Granola

Granola is best for recording customer interviews: video calls, phone calls and in-person meetings, with no bot joining the call. The iPhone and Android apps record in-person interviews, such as site visits and trade shows, and the iPhone app also records phone calls. You can ask across all your meetings, and Claude and ChatGPT can read your notes. It works in 31 languages on desktop and 17 on mobile, and mobile sets one language per meeting, so record mixed-language interviews on desktop. The free plan keeps 30 days of history; Business is $14 for full history.

Reddit Answers

Reddit Answers is best for finding what buyers say in their own words: it is Reddit’s own AI search, and it is free. Copy the threads into your Claude project, ask for the exact phrases buyers repeat, and read the raw quotes yourself. It replaces Gigabrain, which no longer works, and GummySearch, the usual Gigabrain alternative, which closed on November 30, 2025. Check that Reddit Answers is available in your country before you rely on it.

Visualping

Visualping is best for watching competitor pricing and product pages. It alerts you with an AI summary when a page changes. The free plan covers 5 pages, and paid plans start at $14. A weekly Claude scheduled task can check competitor pages and news and summarize them for the team, while Visualping covers the page changes.

What optional add-ons are worth knowing?

Four optional add-ons are worth knowing: F5Bot or Syften for alerts, Ahrefs or Semrush for competitor traffic and keywords, Fathom for video calls, and Notion AI for teams already in Notion.

• F5Bot (free) or Syften (from $29.95) sends ongoing alerts when people mention a topic on Reddit or Hacker News.
• Ahrefs Starter ($29) or Semrush (from $139) shows competitor traffic and keywords. Buy for a month, then cancel.
• Fathom is the alternative to Granola for video calls, with a free plan that has unlimited recordings. Superhuman, the email company, is acquiring it (announced September 14, 2026), so check its free plan after the deal closes. Its in-person support is unclear because its own pages disagree: the iPhone app page and a Fathom help article say different things.
• Notion AI is only for teams already in Notion. Full AI is only on the Business plan, $20 per member, and it has agents and meeting notes.

Which tools cover interviews, surveys, recruiting and market data?

Granola or Plaud Note Pro records interviews, Respondent or User Interviews recruit the people, Google Forms with Gemini or Tally runs surveys, and Perplexity’s premium sources supply market data. Eight more research jobs need a specific tool or feature. The table below gives each one a tool and a price, so the core lineup stays at seven.

When are research platforms worth it?

Research platforms are worth it when you run research at volume, or need a large panel, an advanced survey method or a big archive of public opinion that a $20 assistant can’t give you. Below that, the seven tools above do the job. Here is when five well-known platforms pay off.

For most small B2B teams, three kinds of AI market research tools can wait: AI-moderated interview platforms, enterprise competitor tools, and synthetic users or AI personas.

• AI-moderated interview platforms (Listen Labs, Outset, Strella, Conveo) have no public prices and are built for volume, and because B2B decision-makers are expensive to recruit, these platforms are worth it only past 20 to 30 interviews.
• Enterprise competitor tools (Crayon, Klue, Kompyte) use sales-led pricing and are built for larger teams.
• Synthetic users or AI personas are useful to test an interview guide or form hypotheses, but Nielsen Norman Group warns that they give shallow, overly positive answers and are weakest for niche groups, which is most B2B, so never present them as customer research.

For market research, use ChatGPT for long web research that ends in a finished file: a cited report, a spreadsheet or slides. The ChatGPT Plus plan ($20 a month) includes expanded deep research, projects and ChatGPT Work. Slides are available on Pro, Business and Enterprise. The current chat model is GPT-6, launched October 7, 2026.

Prices and features change monthly. ChatGPT’s live pricing page did not show the Plus price ($20) or the Go price ($8) when this guide was checked, so check both before you pay.

What can ChatGPT do for market research?

ChatGPT can run cited deep research, build documents, spreadsheets and reports, and keep one project per market or client. It also reads your connected apps, analyzes spreadsheets and PDFs, transcribes interview recordings and runs prompts on a schedule. The table shows what each feature does for research and which plan you need.

Custom GPTs retire on December 11, 2026, so do not build one for market research. Create a project instead: it keeps the chats, files and instructions for one market or client.

How do you use ChatGPT for each kind of research?

In ChatGPT, use deep research for market reports, a project for each interview study, and ChatGPT Work for competitor tables and slides.

Market research

• Market or industry report. Run deep research. Before it starts, edit its plan: add your market, region and buyer, and the sources you trust. Export the report to Word and check the key numbers.
• Market sizing from data. Upload the data (an export, an analyst table) and ask for the calculation step by step, with the formula shown.

Customer research

• Interview analysis. Create a project per study. Upload transcripts or audio. Run the extraction prompts below one at a time, in the same project, so it keeps the context.
• Budget and buying triggers. Ask across all transcripts in the project: “Which interviews mention budget, who owns it, and what triggered the search?” Ask for the quote and the interview name for each.

Competitor research

• Competitor table. Give it the URLs and use ChatGPT Work to build a spreadsheet with one row per competitor and a link per cell.
• Pricing page checks. Use the built-in browser in ChatGPT Work to open each pricing page and record plans and prices into the sheet.

Strategy

• From research to slides. Ask ChatGPT Work to turn the report into a short deck for the team (slides on Pro, Business and Enterprise).

How do you protect your data in ChatGPT?

To keep your ChatGPT chats out of model training, go to Settings > Data controls and turn off the “Improve the model for everyone” setting. Your history stays. The setting is on by default for Free, Go, Plus and Pro. Temporary chats are not used for training. Business and Enterprise don’t train on your data by default. Note: a thumbs up or down can still send that conversation for training.

For market research, use Claude when you give it your own material (transcripts, notes, reviews) and want the same method applied every time: one project per client or market keeps the files, instructions and memory together. Claude Pro ($20 a month) gives you Research, projects with memory, skills and the browser agent (Claude in Chrome).

What can Claude do for market research?

Claude can search the web with citations, run Claude Research across the web and your connected apps, and keep one project per client or market. It can also save your method as a skill, create and analyze files, browse sites for you and run tasks on a schedule.

How do you use Claude for each kind of research?

You use Claude for each kind of research by working in a project that holds your own material. Research builds the market brief and the competitor table, uploaded transcripts feed the customer research, and a skeptical second opinion tests your strategy.

Market research

• Get up to speed on a new market in a day. Create a project for the market. Add the client’s deck and website notes. Turn on Research and ask for a cited brief on size, players, buyers and risks. Ask it to mark which claims come from a source and which are its reasoning.
• Find market stats. Use Research, then open the three most important sources yourself before you quote a number.

Customer research

• Extract pains, jobs and outcomes from interviews. Upload transcripts (or connect Granola, which Claude can read through its connector) to a project. Run the extraction prompts below one at a time.
• Make it repeatable. Once a prompt works, ask Claude to turn it into a skill. Next time, uploading a transcript is enough.
• Find patterns across interviews. Put all transcripts in one project and ask what repeats, with quotes and the interview each quote came from.

Competitor research

• Build a competitor table. Use Research with the competitor URLs; ask for an Excel file with one row per competitor and a source for each cell.
• Read reviews in bulk. Paste or upload a competitor’s reviews from G2 or Capterra and ask for themes with counts. Check each site’s terms first. The rules and the full method are in the guide to analyzing competitor reviews with AI below.
• Watch competitors. Set a weekly scheduled task to check their pricing pages and summarize changes.

Strategy

• A second brain that disagrees. In the project, ask Claude to argue against your positioning, as a skeptical buyer would.

How do you protect your data in Claude?

To protect your data in Claude, check the Model Improvement setting in Settings > Privacy, and for client interviews use a Team plan or an incognito chat and remove names first. On Free, Pro and Max, Claude trains on your chats only if you allow it. Incognito chats are never used for training. Team and Enterprise don’t train on your data by default.

To calculate TAM, SAM and SOM with AI, count your buyers from the bottom up with free government data, let AI do the math in a spreadsheet you can see, and use analyst numbers only as a sanity check. That’s the method investors trust: Pear VC’s market sizing guide calls bottom-up “more convincing” and says top-down figures are only for checking the order of magnitude.

What do TAM, SAM and SOM mean?

TAM, SAM and SOM are three numbers that size a market: the total market (TAM), the market you can serve (SAM) and the market you can win (SOM).

Which tools do you need to size a market with AI?

To size a market with AI you need Claude or ChatGPT for the industry codes, free Census and Eurostat data for the company counts, a sales database to cross-check them, Claude for the math, and deep research for the top-down check.

Skip the “AI TAM calculator” websites: most are lead pages for data vendors and hide their assumptions. Skip Semrush’s “Market Size (TAM)” too: it is website traffic, not dollars or buyers.

How do you size a market step by step?

You size a market in eight steps: write the buyer in one sentence, find the industry codes, count the companies, set the yearly price per company, build the model, cut down to SAM, estimate SOM from your sales capacity, and sanity check the numbers top-down.

1. Write the buyer in one sentence. Industry, company size, country, and the role that buys. Take it from your ideal customer profile.

2. Find the industry codes. Ask Claude for the matching NAICS codes (prompt below). Check each code’s description on census.gov before using it.

3. Count the companies. Download the Census table for those codes, filtered by employee size. Upload it to a Claude project. Then run the same filters in Sales Navigator or Apollo. If the two counts are far apart, find out why before you go on.

4. Set the yearly price per company. Use your own average contract value from closed deals. No deals yet? Use the entry price on 3 competitors’ pricing pages, and say so.

5. Build the model in Claude. One tab per number, every assumption in its own cell, formulas visible (prompt below). TAM = companies × yearly price.

6. Cut down to SAM. Remove the countries, company sizes and segments you can’t serve or reach today. Write each cut and why.

7. Estimate SOM from your sales capacity. Work out how many deals your team can close each year over 3 years (reps × deals per rep, or your own pipeline numbers). Show low, middle and high. If SOM is more than a small share of SAM, question it.

8. Sanity check top-down. Run deep research for analyst figures on the same market (prompt below). Compare the order of magnitude, not the exact number. Open every source.

Which prompts size a market?

Three prompts size a market: one finds the industry codes (step 2), one builds the Excel model (step 5) and one checks the result against published estimates (step 8).

Industry codes

Use this in step 2: replace the bracketed text with your one-sentence buyer from step 1.

My buyers are: [one-sentence buyer]. List the 2022 NAICS codes (6 digits) that best match them. For each: the code, its official title, why it fits, and what it includes that is NOT my buyer. Flag codes that only partly fit.

Build the model

Use this in step 5, in the Claude project that holds your Census table and your notes.

Using the attached Census table and my notes in this project, build an Excel file with three tabs: TAM, SAM, SOM. Put every assumption in its own labeled cell (company counts by size band, yearly price, the cuts for SAM, deals per year for SOM). Use formulas, not typed results. Add a tab listing every source with its link and year. Show SOM as low, middle and high.

Top-down check

Use this in step 8 with deep research: replace the bracketed text with your market and region.

Find published estimates of the size of [market] in [region]. For each: the figure, the year, who published it, how they define the market, and the link. Mark estimates that come from a press release by a report seller. Do not give me a figure you can’t link to.

What should you check before you use the numbers?

Before you use your market-size numbers, make sure each has a source and a year, check how any market-size headline defines the market, find each figure AI gave you on the page it links to, and write your assumptions on the slide.

• Every number has a source and a year. Census business data runs a few years behind (2022 is the latest by company size). Say so.
• Don’t trust a market-size headline. Big round figures in press releases often come from report sellers and get repeated until they look like fact. Check how the market is defined.
• AI can invent figures. Open each link and find the number on the page.
• Write your assumptions on the slide. There is no single definition of market size; the assumptions are what an investor checks.

To analyze competitor reviews with AI, collect the reviews into one spreadsheet, let AI propose themes, then have it count them with code, not by impression. The result is a ranked list of your competitors’ weak spots, with real quotes you can trust.

What are the rules for collecting reviews?

The rules for collecting competitor reviews are strict: G2, Capterra, TrustRadius, Gartner Peer Insights and Trustpilot all forbid scraping in their terms, even for public pages. Capterra is now owned by G2. G2’s and Capterra’s terms (updated in 2026) also forbid using their content with AI or machine learning.

None of these sites offers a bulk export of a competitor’s reviews. App stores only give you your own app’s reviews through their official APIs. Check the terms of each site before you collect anything, and read G2’s community guidelines as well.

Which tools collect and analyze reviews?

Mining competitor reviews takes three kinds of tools: one to collect the reviews, Claude or ChatGPT to find the themes and count them, and Gemini Notebook to ask questions across them. The safer ways to collect are your own G2 data (if you are a G2 vendor), app store tools and copying a few dozen by hand. Bulk collection comes with a warning: the review sites forbid scraping, so read their terms first.

Skip: enterprise feedback platforms (Thematic, Unwrap, Chattermill, Enterpret) start around $24,000 a year. Kraftful was bought by Amplitude and Viable appears to be gone.

How do you analyze competitor reviews step by step?

You analyze competitor reviews in seven steps: pick 3 to 5 competitors and one source each, collect at least 100 reviews per competitor, find the themes on a sample, tag every review with code, check the tags by hand, pull the quotes with their IDs, and turn the repeating complaints into decisions.

1. Pick 3 to 5 competitors and one source each. G2 for software, the app stores for apps, Reddit for what people say when no one asks them to review. Read each site’s terms before you collect: G2’s and Capterra’s forbid using their content with AI, so on those sites start from your own G2 data.

2. Collect at least 100 reviews per competitor. Use reviews from the last 24 months. One row per review: an ID, date, rating, reviewer role, company size, pros, cons, and whether it is marked “Incentivized.”

3. Find the themes on a sample. Upload the file to a Claude project and run the theme prompt on 50 reviews. Edit the list until it uses your buyers’ words. 8 to 15 themes is enough.

4. Tag every review with code. Run the tagging prompt. Claude writes code that adds one column per theme and counts them by competitor, rating and company size.

5. Check by hand. Read 30 to 50 tagged rows. Fix the theme definitions where the tags are wrong, and run it again.

6. Pull the quotes with their IDs. Every quote must have a review ID that code matches back to the file. Don’t accept a quote that isn’t in the file. The gaps prompt below checks this with code.

7. Turn it into decisions. The complaints that repeat across competitors are the gaps you can own. They go into the competitor landscape (playbook step 4) and your positioning angles (playbook step 6).

Which prompts analyze reviews?

Three prompts analyze competitor reviews: one finds the themes in a sample, one tags every review with code and counts the themes, and one lists the top complaints with exact quotes.

Find the themes

Use this in step 3, on a sample of 50 reviews: replace the bracketed text with your competitors, the buyer role and the company size.

Here are 50 reviews of [competitors] by [buyer role] at [company size]. Propose 8 to 15 themes for what reviewers like and dislike. For each theme: a short name in the reviewers’ own words, a one-line definition, and two review IDs that show it. Keep likes and dislikes as separate themes.

Tag and count

Use this in step 4, in the Claude project that holds your themes, with the review file attached.

Using the themes in this project, write and run code that tags every review in the attached file (a review can have several themes). Add one column per theme. Then count: each theme by competitor, by star rating and by company size, with and without reviews marked Incentivized. Give me an Excel file with the tagged reviews and a summary tab.

Find the gaps

Use this in steps 6 and 7, in the same project: it needs your product notes there to say which complaints your product solves.

From the counts, list the 5 complaints that come up most across all competitors, with the count, the share of reviews, and 2 exact quotes each with their review ID. Check with code that each quote appears word for word in the file. Then say which of these our product solves, based on the product notes in this project.

What should you check before you use the findings?

Before you use the findings from a review analysis, check four things: whether incentivized reviews skew the counts, whether happy customers are over-represented, whether a complaint is already fixed, and whether every quote matches a review ID.

• Incentivized reviews. G2 marks reviews where the reviewer got a gift. Compare results with and without them.
• Happy customers are over-represented. Vendors invite their fans to review, so the complaints are the more useful signal.
• Old complaints may be fixed. Check the competitor’s release notes before you build an angle on a weakness.
• AI invents quotes. A 2025 study found chatbots made things up in 60% of answers about product reviews. Only use quotes matched to a review ID.

A good AI market research prompt has four parts: a context block, one clear task, the output shape you want and a rule on sources. Every prompt in this section follows that pattern. The context block goes into your project instructions once, so each prompt can say “using the context” instead of repeating your market, your buyer and your hypotheses.

Role-play lines such as “You are a 20+ year expert,” “Do you understand?” and a $100 tip no longer improve answers. Context does. Run the prompts in a project in ChatGPT or Claude.

This section gives you the context block, 16 prompts for market, customer and competitor research, and a follow-up prompt for deeper answers.

Step 0: the context block

The context block is a short note that tells the AI your product, market, buyer, hypotheses and the decision you will make, and you put it in your project instructions once. Use one project per market or client, so every chat in it starts informed. Fill in the [square brackets] with your own details, here and in every prompt that has them. The two rules at the bottom make the tool label each claim as SOURCE, with a link, or as REASONING, and say so when it does not know instead of guessing.

Context
• Product or idea: [one or two sentences]
• Market and region: [e.g. incident management software, US]
• Buyer and end user: [role, company type, company size]
• What we already believe: [your hypotheses]
• What I will decide with this research: [the decision]

Rules
• Mark each claim as SOURCE (with link) or REASONING.
• If you don’t know, say so. Don’t fill gaps with guesses.

Prompts for market research

Market research prompts ask AI for market stats, research sources, barriers to adoption, market maturity and the segments to win first. These are the use cases for this kind of research:

• Discovering market stats
• Discovering potential research sources
• Discovering market category differences and market perceptions
• Discovering barriers to product success (risks of adopting your product category)
• Discovering the perceived market risks of adopting a tech or product
• Understanding customer’s behavioral change and the learning curve of adopting a new tech
• Researching inside a PDF (with ChatGPT)
• Understanding market maturity (stage of the innovation adoption curve)
• Exploring opportunities and potential targets
• Understanding challenges by [AUDIENCE] in [CATEGORY]
• Discovering problems for a target market from the perspective of the client
• Discovering problems your tech/product can solve from a product perspective
• Discovering potential segments and applications to target
• Ranking segments to win first

Six ready-to-paste prompts follow.

Market stats

Use this prompt to find the latest market figures, then open the three most important sources yourself before you quote a number.

Using the context, find the latest figures on the size and growth of this market and the markets next to it. One bullet per stat: a short bold title, the number, the year, and the link. Prefer analyst reports, public filings and government data. Flag any figure older than two years.

Sources to read first

Use this prompt to find the best places to learn about your product category.

List the 20 best places to learn about this product category: analyst reports, industry associations, communities (Reddit, Slack groups, forums), newsletters and podcasts. For each: link, what I’ll learn there, and how current it is.

Barriers to adoption

Use this prompt to see what stops your buyer from switching; goal 4 runs it in ChatGPT deep research.

What stops buyers in this market from adopting this kind of product? For each barrier: who raises it (buyer or end user), how it hurts their goals or daily work, and one piece of evidence (link) that it is real.

Market maturity

Use this prompt to see where your product category is on the adoption curve today (see also the Crossing the Chasm guide).

Where is this product category on the adoption curve today? Give your conclusion first, then the reasons. Describe the buyers who are buying now: what they value, how much risk they accept, and what proof they need. Link each reason to a source.

Segments and use cases

Use this prompt to list 5 to 10 segments from your product and best clients, as in goal 3.

List 5 to 10 segments that could use this product. A segment is a type of company plus the buyer’s role. For each, list the processes where the product could solve a problem, and the problem in the buyer’s words. Use only the context and the files in this project.

Rank the segments

Run this prompt right after the segments prompt, so it scores the segments you just listed and names the one to win first.

Score the segments above from 1 to 5 on: how painful the problem is, budget, how easy they are to reach, how few competitors serve them, and whether a win there helps us win the next segment. Show the table, then name the one segment to win first and why.

Prompts for customer research

Customer research prompts turn your interview transcripts into pains, jobs, outcomes, patterns across interviews, the buyer’s journey and the workflow change, so attach your transcripts before you run them. These are the use cases for this kind of research:

• Extracting pain points from customer/prospect interviews
• Extracting jobs to be done from customer/prospect interviews
• Extracting outcomes from customer/prospect interviews
• Summarizing patterns from multiple market and customer interviews
• Understanding budgets (is there a budget allocated to solve the problem my product solves?)
• Exploring pain points starting with an audience and category of product
• Assisting in a discussion of potential targets
• Discovering the customer’s journey
• Exploring the economic penalties for an audience for not adopting a specific type of product/solution
• Discovering the end-user’s workflow impact of implementing a given solution/tech
• Customer scenario discovery (before, after, bridge)
• Discovering buying trigger events

Six ready-to-paste prompts follow. Run the pains, jobs and outcomes prompts one at a time in the same project, so the tool keeps the context.

Pain points

Use this prompt to pull the pains out of one interview, with no jobs or outcomes mixed in.

From this interview with [persona], list the pains, struggles and frustrations. Pains only, no jobs or outcomes. For each: a short title, a direct quote, and how often it came up.

Jobs to be done

Use this prompt to list the jobs in one interview, in order of importance, in the Ulwick style (see also the jobs to be done example).

From this interview, list the jobs to be done, in order of importance to the interviewee. Write each as Verb + Object + Clarifier. Add the quote that supports it.

Desired outcomes

Use this prompt to list the outcomes one interviewee wants, with no pains or jobs mixed in.

From this interview, list the outcomes the interviewee wants. Outcomes only. Write each as Minimize or Maximize + time, number, frequency, likelihood or ability + “to” + object. Add the supporting quote.

Patterns across interviews

Use this prompt once all your transcripts are in one project, to see what repeats across interviews.

Across all transcripts in this project, what pains, jobs and buying triggers repeat? Show a table: theme, how many interviews mention it, the strongest quote, and which interviews. List anything mentioned only once separately.

Customer journey

Use this prompt to map the buyer’s journey for one use case and see where it hurts.

Map the buyer’s journey for [use case], step by step. For each step: what they are trying to get done, the tools they use, and where it hurts. Mark which steps come from our interviews and which are your reasoning.

Workflow change

Use this prompt to see which daily tasks change for the end user if they adopt your product.

For the end users in the context, which daily tasks change if they adopt this product? For each: what they do today, what they would do instead, and how hard the switch is (low, medium, high).

Prompts for competitor research

Competitor research prompts build a competitor table from a URL, show why competitors win, mine reviews for gaps and prepare you for a head-to-head deal. These are the use cases for this kind of research:

• Discovering competitor (or your company) whitespaces and weaknesses from reviews
• Discovering competitors in existing categories
• Discovering competitors in categories in formation or unknown
• Create a competitor table by providing a product/company URL
• Discovering perceptions about a competitor (or your company)
• Discover why competitors win
• Creating a SWOT analysis
• Getting an AI opinion on your product/company going head to head against XYZ competitor
• Understanding the market perception of your company or a competitor

Four ready-to-paste prompts follow.

Competitor table from a URL

Use this prompt to build a competitor table from a product or company URL, with a source link for each cell.

Here is our website: [URL]. Find the 6 to 10 closest competitors. Build a table: name, URL, who they sell to, value proposition, key differences, entry price, and a source link for each cell. Export it as a spreadsheet.

Why competitors win

Run this prompt after the competitor table, to see why buyers choose each competitor in it.

For each competitor in the table, why do buyers choose them? Use reviews (G2, Capterra), case studies and their own sales pages. One line per competitor, with links.

Review mining

Use this prompt once you have copied a competitor’s reviews into a Claude or ChatGPT project; the mini guide on how to analyze competitor reviews with AI has the full method.

Here are [number] reviews of [competitor]. Group the complaints and the praise into themes. Show each theme with a count and one quote. End with the three gaps we could own.

Head to head

Use this prompt before a deal against a named competitor, to see why the buyer might pick them and what you would need to show to win.

We are up against [competitor] in a deal with [buyer]. Argue the buyer’s side: why they might pick the competitor over us. Then list what we would need to show to win.

What follow-up prompt gets deeper answers?

The follow-up prompt that gets deeper answers tells the AI its first answer is too general, asks it to go one level deeper for one segment with specific situations, numbers and quotes, and asks it to say so where it has no evidence. Use it whenever an answer comes back too general.

This is too general. Go one level deeper for [segment]: name specific situations, numbers and quotes. Where you have no evidence, say so.

Predictable Innovation has completely embedded AI in its market research workflow to provide clients with quicker results while keeping the same high quality. We’ve found 25 to 40% time savings thanks to using Artificial Intelligence (AI) for market research and other use cases. We’ve identified and documented 67 use cases in which AI can be used for customer, competitor, and market research, and some of them are detailed in this article.

The role of AI in market research has been transformative for our business and competitiveness, providing our team with different benefits:

Gaining analysis efficiency

We were analyzing tons of Reddit posts, call transcripts, reading entire analysts’ PDFs, or hundreds of G2 competitor feedback answers.

That’s gone, using AI-powered market research. AI makes data collection from these sources almost automatic. It is a great tool for finding patterns, summarizing information, and extracting data from text or videos. We use all this to gather market intelligence faster.

Gaining a second brain and new perspectives

AI gives us fresh angles for product positioning and go-to-market strategies that are not biased by our beliefs.

Getting broader research insights

AI tools can identify sources and reach internet corners we could not or did not have the time. We’ve gained more research findings, insights, understanding, and angles about the market, customers, and competitors we analyze.

Taking off research projects at light speed

With traditional market research methods, we had to conduct secondary internet research and analyze interviews. It would take around two weeks to warm up a market analysis or strategy project.

Now, by using AI-powered market research tools, we get up to speed on new client projects in just one day by using AI tools for market research.

Faster market feedback loops & strategy development

Given the speed at which we can now analyze the information, we can deliver faster market testing and feedback loops, leading to a competitive edge.

Market learning speed is often an advantage.

Overcoming the blank-page syndrome

We won’t have to feel the pain of starting from 0 anymore to fill in our templates and create strategies. AI is helping us with a fresh starting point.

We’ve measured the impact of integrating AI for market research and other use cases in our workflow. Here’s what we’ve found:

25 to 40% less time on research

Depending on the project’s complexity, we’ve measured 25 to 40% working time savings in market/customer research, innovation, positioning strategy, messaging, and sales tasks.

P&L IMPACT: All that goes directly into Predictable Innovation’s P&L bottom line.

x2 productivity

In strategy projects for clients from the kick-off to deliverables. An average project was three months; with AI, it takes 6 to 8 weeks. Most of the productivity gains are coming from AI market research use cases.

STRATEGIC IMPACT: Speed-to-delivery is now a key feature of Predictable Innovation. We now deliver high quality with more speed to results.

New research eyes and strategic angles

We spent tens of hours analyzing Reddit posts, interview transcripts, analysts’ PDFs, google results, and G2 user interviews. That’s now gone with all the AI tools in this article.

TEAM IMPACT: The AI is working as their team assistant to provide angles for their strategy team. These are new strategic angles that are not biased by preconceived human beliefs.

The best practices for using AI in market research are to use one project per market or client, give AI rich context, check what it says, never skip real customer interviews, protect sensitive data, ask follow-up questions and save what works.

1. Use one project per market or client

Put the context block in the project instructions so every chat starts informed.

2. The more specific you are, the more specific the answer

AI is as good as the question you’re asking and the context you’re providing. Be clear, specific, and relevant: give it the context block, the decision you will make, and the output shape you want.

3. AI is your assistant; you’re the brain

Let AI provide you with data, angles, and analysis. But you and your team MUST double-check it and make informed decisions. Ask the tool to mark each claim as a source or as its own reasoning, and open the key sources yourself before you quote a number.

4. AI is NOT a replacement for customer or market interviews

What you’ll find in a single customer interview, or even an interview with your sales team, will be way more detailed, specific, and relevant than what you’ll find with AI research over the internet. Synthetic personas are for preparing interviews, not replacing them.

5. Protect sensitive data

In ChatGPT, turn off “Improve the model for everyone” under Settings > Data controls. In Claude, don’t allow Model Improvement under Settings > Privacy. Remove names from transcripts before you upload them.

6. Follow-up questions are a must to discover more specific answers

Think of AI as one more team member: you won’t just take the first answer and use it, so get curious, keep asking, and go deeper. Run the same question in two tools (for example ChatGPT deep research and Gemini Deep Research) and compare the answers. Where they disagree, check by hand.

7. Push for a more specific answer

If the answer is not what you expect, make your prompt more precise, specific, and relevant. To push further, use the follow-up prompt. It asks the tool to go one level deeper with specific situations, numbers and quotes, and to say so where it has no evidence.

8. Save what works

Turn a prompt that works into a Claude skill, or keep it in a ChatGPT project, so the team runs the same method every time.

The main challenges of using AI in market research are data quality, transparency, job displacement, AI bias, the rapid pace of AI development and the change management it needs.

Data quality

AI models are only as good as the data they are trained with, so incomplete or inaccurate data gives unreliable results. Always double-check the results and find the sources of truth: check sources, dates and whether the number is for your region. The best practices above add more checks, such as marking claims as sources or reasoning and comparing two tools.

Transparency

AI is often called a “black box”: it is difficult to understand how these models make their outputs, which can lead to mistrust and resistance from stakeholders. Deep research in ChatGPT and Research mode in Claude show their sources and steps, so read them. Start with pilots and showcase the value of AI for market research.

Job displacement

Job displacement has no easy solution; the best you can do is use AI to augment your team’s capabilities rather than replace them. That frees your team to focus on more complex and strategic tasks while they use AI for the use cases in this guide. Artificial intelligence is the assistant; human intelligence is the decision-maker.

AI bias

If the data used to train an AI is biased, its predictions and analyses will be biased too. That can lead to inaccurate market research results and potentially harmful decisions. Regular audits of AI systems can help identify and correct biases.

Rapid pace of AI development

AI technology constantly evolves, making it difficult for organizations to stay up-to-date with the latest tools and techniques. Three of the tools this guide featured in 2024 no longer work, which proves the point. The solution is to invest in continuous learning and development: attend industry conferences, take online courses, partner with AI experts or subscribe to an AI newsletter like the Neuron.

Change management needs

Implementing AI may require significant changes to workflows and processes, and employees who are comfortable with existing research approaches may resist. To overcome this, start with small pilots, communicate the benefits of AI clearly, and provide training and support during the transition. Involving employees and addressing their concerns also helps ensure a smooth transition.

These challenges can be overcome with the right strategies and approaches.

By focusing on data quality, transparency, ethical considerations and continuous learning, you can successfully use AI to improve your market research.

Watch this video on how Predictable Innovation uses AI for one concrete use case, the ideal customer profile:

Prices are in US dollars per month unless noted, checked on November 2, 2026. Prices and features change often, so check the vendor’s page before you buy.

ChatGPT: Pricing · Deep research · Release notes · Projects · Connected apps · Data controls · ChatGPT Record

Claude: Pricing · Research · Web search · Projects · Skills · Custom skills · Connectors · Files and analysis · Claude in Chrome · Scheduled tasks · Training and privacy

Other tools: Gemini plans · Gemini Notebook rename · Perplexity Pro · Perplexity plans · Fathom pricing · Superhuman acquires Fathom · Grain pricing · Granola pricing · Notion pricing · Visualping pricing · F5Bot · Syften pricing · Ahrefs pricing · Semrush pricing

Interviews, surveys and recruiting: Granola mobile app · Granola languages · Granola chat across meetings · Google Meet notes in person · Gemini Notebook sources · Fathom iPhone app · Respondent · User Interviews · Sales Navigator · Clay · Gemini in Forms · Tally · Apify · Perplexity premium sources · Semrush Market Explorer · Plaud · Plaud Note Pro review

Market sizing and review mining: Pear VC market sizing guide · Census SUSB tables · Census County Business Patterns · NAICS · Eurostat business statistics · Apollo pricing · Statista plans · IBISWorld report price · Semrush Traffic & Market · Claude for Excel · IPVM on report sellers · G2 terms · Capterra terms · TrustRadius terms · G2 buys Capterra · G2 community guidelines · Apify review scrapers · Bright Data G2 dataset · Appbot plans · Thematic pricing · Amplitude buys Kraftful · Study on AI review summaries

Research platforms (checked on November 3, 2026): Quantilope pricing · Qualtrics market research · Attest pricing · Outset pricing · Brandwatch

Shutdowns and warnings: Aomni site · GummySearch closing notice · Essense listing · Reddit Answers growth · Nielsen Norman Group on synthetic users

Written by

Jose M Bermejo

Strategic growth partner, coach and keynote speaker for B2B founders, owners and CEOs

15+ years selling B2B tech, first as a chief revenue officer who grew revenue 120%. He ran Microsoft partnerships, trained at IESE, Harvard Business School and MIT Sloan, and wrote B2B Positioning Strategy, with contributions from Warren Schirtzinger.

More about JoseLinkedIn