Traditional market research takes days: competitor analysis, customer persona development, TAM sizing, SWOT breakdowns. With the right prompts and workflow, you can compress that into a few hours — and get output that's actually useful for decision-making, not just filler for a slide deck. Here's the exact workflow.
The important nuance is that AI does not replace judgment. It gives you a faster first pass on structure, language, and pattern recognition so you can spend your real energy validating assumptions and making decisions. That is a much better use of your time than manually drafting frameworks you already know by heart.
Before diving into prompts, a quick calibration: ChatGPT is excellent at synthesis, structuring frameworks, generating hypotheses, and drafting customer personas based on patterns in its training data. It's not a real-time data source — it doesn't have yesterday's revenue figures for a private competitor or live search volume data.
The workflow below uses AI for what it's genuinely good at, and points you to the right tools (Google Trends, SEMrush, Crunchbase, Reddit) for what it's not. That combination is where the real leverage is.
If you treat AI output as confirmed truth, you will make sloppy decisions. If you treat it as an analyst who is fast, tireless, and occasionally overconfident, you will get much more value from it. The job is to use it for acceleration, then verify anything material before it shapes roadmap or budget choices.
Most market research fails because the scope is fuzzy from the start. Use this prompt to force precision:
I'm researching the market for [product/service idea]. Help me define this market precisely by answering: 1. What is the narrowest useful definition of this market? 2. What are the adjacent markets I should be aware of? 3. Who are the primary customer segments within this market? 4. What problem does each segment have that creates the purchase trigger? 5. What are the existing alternatives customers use today (not just direct competitors)? Be specific — avoid generic consulting-speak. Give me concrete examples.
The output gives you a market map you can validate with data — much faster than starting with a blank page.
A helpful follow-up is to ask the model what this market is commonly confused with and why. That usually reveals where your messaging may drift later. If your category definition is muddy at the research stage, it will almost certainly be muddy on the website and in the sales process too.
Once you have your market defined, map the competitive landscape:
I'm building a [product type] that [core value proposition]. My target customer is [customer segment]. Map the competitive landscape: 1. List the top 6-8 direct competitors (name, positioning, rough price point if known) 2. List 3-4 indirect competitors or substitutes 3. For each direct competitor, note: their core differentiation, their apparent weakness, and who their ideal customer is 4. Where are the gaps in this market that none of them are serving well? Base this on the [market/industry] space as of your training data.
Then validate: search each competitor name in Crunchbase (funding, team size), pull their traffic in SimilarWeb or Ahrefs, and read their 1-star reviews on G2 or Trustpilot. The reviews are gold — they tell you exactly what customers hate about existing solutions.
This is also the stage where AI can help you avoid lazy positioning. If every competitor says "save time," "boost productivity," or "streamline workflows," ask the model to surface the more specific language customers use when describing why they switch. That language often gives you a better differentiator than anything you would invent in a brainstorm.
If competitive content analysis is part of your process, pair this with our guide on AI for SEO optimization so your market research also informs your search strategy instead of living in a separate doc no one revisits.
Generic personas ("marketing manager, 35, likes efficiency") are useless. Use this prompt to generate ones that actually inform product decisions:
Create 3 detailed customer personas for [product] targeting [market]. For each persona, include: - Name, role, company size, industry - Their day-to-day workflow where this product fits - The specific problem that would make them buy today (the trigger event) - What objections they'd raise on a sales call - How they currently solve this problem without your product - What a home-run outcome looks like for them 90 days after purchase - Where they spend time online (communities, publications, social platforms) - What they'd type into Google when they start looking for a solution Make these feel like real people, not marketing composites.
Once the personas come back, pressure-test them with one more prompt: "What would make this persona ignore my product entirely?" That single question is useful because it surfaces the cases where you are chasing a segment that sounds plausible on paper but will never convert in practice.
Market size estimates need to be built from first principles — not just pulled from a report. Use AI to build the reasoning:
Help me size the market for [product/service] using a bottom-up approach. My target customer: [description] Geography: [US / Global / specific region] Rough price point: [annual or monthly] Walk me through: 1. How many potential buyers exist? (give your reasoning, not just a number) 2. What % realistically would consider this category? (SAM) 3. What's a realistic 3-year capture rate given the competitive dynamics? (SOM) 4. What assumptions am I making that could be wrong by 2x in either direction? 5. What publicly available data sources should I check to validate these estimates?
The point isn't to get a perfect number — it's to understand your assumptions well enough to defend them to an investor or executive.
That distinction matters. Bad market sizing is usually not wrong because the math was imperfect; it is wrong because the assumptions were hidden. AI helps by making the assumptions visible quickly, which makes your next round of validation far more targeted.
If you've done customer interviews, collected Reddit threads, or gathered review data — AI is exceptional at synthesizing large volumes of qualitative text into patterns.
I've collected the following customer feedback / interview notes / forum posts about [problem space]. Analyze this for: 1. Top 5 recurring pain points (ranked by frequency) 2. Exact language customers use to describe the problem (quote directly) 3. Jobs-to-be-done: what are they actually trying to accomplish? 4. Any surprising insights that contradict common assumptions about this market? 5. Unanswered questions I should follow up on in the next round of research Raw data: [paste your notes, Reddit threads, review excerpts, interview summaries]
This prompt works particularly well when you paste in 10-20 Reddit comments from relevant subreddits, or 20-30 G2 reviews from a competitor. The synthesis usually surfaces patterns that would take hours to find manually.
Be careful here not to lose the original language. If three users describe the same pain in blunt, memorable words, keep those words. They are often more valuable than the cleaned-up summary because they become better copy for landing pages, ads, onboarding flows, and sales scripts.
The final step is converting your research into a decision-ready brief. Use this to wrap everything up:
Here's my market research so far: [paste your findings] Write a one-page market research brief that includes: 1. Market definition and size (with confidence level) 2. Top 3 customer segments and their priority ranking 3. Competitive white space — where there's room to win 4. Top 3 risks or assumptions that need validation before proceeding 5. Recommended next steps (what to build, test, or learn first) Write it for an executive audience — direct, no fluff, actionable.
This brief becomes the artifact you share with stakeholders, pitch to investors, or use to align your team before building anything.
If your team runs this process regularly, store the finished briefs in Notion or a similar knowledge base so patterns compound over time. Market research gets much more powerful when each new project can borrow from past competitor notes, persona language, and validation questions instead of starting cold every time.
For a broader workflow that combines research, synthesis, and operational follow-through, our using Claude for research guide is the natural next step.
💡 Want to automate ongoing market monitoring? Build a Make.com scenario that tracks competitor blog posts, Reddit mentions, and news articles weekly — and sends you a digest summarized by AI. It takes about an hour to set up and runs itself from there. See all recommended tools →
Practical prompts and automation ideas — no fluff.