01 · Reality check
What Claude actually does well here
Good at
- Structuring a vague question into a researchable set of sub-questions
- Reading long competitor documents, filings and reports and pulling out patterns
- Synthesising several sources into a coherent picture with the tensions noted
- Designing survey and interview questions that avoid leading the respondent
- Challenging your assumptions about the market when instructed to
Not the right tool for
- Producing reliable market-size figures, which it can invent convincingly
- Knowing recent events unless web search is on and it actually found them
- Primary research: it cannot talk to your customers
- Distinguishing a credible source from a content-farm article without help
- Judging whether a market is right for you, which needs your context
02 · The method
Step by step
- 1
Turn web search on
It is available on every plan including Free, but it is not enabled by default. Toggle it from the chat input, otherwise Claude answers from training data alone.
- 2
Break the question down first
Ask it to turn your broad question into specific sub-questions, each answerable with evidence. Vague research questions produce vague, useless reports.
- 3
Demand a source for every claim
Instruct it to cite the source for each factual statement and to say plainly when it does not have one. This single instruction changes the quality of the output more than anything else.
- 4
Feed it primary documents
Upload competitor reports, filings and pricing pages. Analysis of documents you supply is far more reliable than asking what it knows about a market.
- 5
Verify the numbers that matter
Open the source for every figure you plan to use in a decision or a deck. Market-size numbers are where AI research most often goes wrong, and where being wrong is most expensive.
- 6
Have it argue against the conclusion
Ask for the strongest case that your read of the market is wrong. This surfaces the assumptions you did not know you were making.
03 · Use this now
Copy-paste prompt for structured market research
Act as a rigorous research analyst. My research question: [state it]. My context: [what decision this informs]. First, break this into 5 specific sub-questions that can each be answered with evidence, and tell me which are answerable from public sources and which need primary research. Then research the answerable ones. Rules: cite a source for every factual claim; state plainly when you do not have a source rather than estimating; flag any figure you are uncertain about; distinguish clearly between what you found and what you are inferring. Rank the sources you used by credibility and tell me which claims I should verify myself before relying on them.
04 · Avoid these
Common mistakes
- Accepting market-size figures without opening the source. This is the most common expensive error.
- Forgetting web search is off by default, so it answers from training data and you assume it searched.
- Treating a well-formatted report as a well-evidenced one. Structure is not accuracy.
- Using it as a substitute for talking to actual customers, which it cannot do.
- Letting it confirm what you already believe rather than instructing it to challenge you.
05 · Questions
Frequently asked questions
Can Claude do market research?
It is genuinely useful for structuring the question, reading long competitor documents, and synthesising sources into a coherent picture. It is not a reliable source of market statistics on its own. The workflow that works is Claude for structure and synthesis, with you verifying every figure at its original source.
Does Claude search the web?
Yes, on all plans including Free, but it is not on by default. You enable it from the chat input. Paid plans also get Research, an agentic mode that runs multiple searches for deeper work, and it requires web search to be enabled first.
Can I trust the market size figures Claude gives me?
Not without checking. Market-size numbers are exactly where AI research goes wrong, because plausible figures are easy to generate and hard to spot as wrong. Instruct it to cite a source for every number, then open the sources for any figure you will actually use in a decision.
What is Claude Research mode?
It is an agentic research feature that runs multiple searches and works through a question in more depth than a single response. Anthropic makes it available on paid plans, and it requires web search to be turned on. It is best for questions needing several angles rather than one lookup.