- Your first Project and file uploads
- The major AI families to try the workflow
- A small daily message allowance
Research with AI
Research with AI without believing the first confident answer.
AI is a fast researcher and an unreliable witness. It will give you a market size, a competitor's price and a regulation's effective date in one fluent paragraph, and one of the three will be wrong. Good research with AI is a method: ask a model that cites sources, have another check it, put the evidence in a Project so it can be questioned later, and verify anything you will act on. This page is that method for market, competitor and product research.
Free to start. Sourced research and the latest models are on paid plans.
The method
Five rules that make AI research trustworthy
Ask for sources, then open them
Use a model that cites (Perplexity is built for it; research modes in others). Open every source you will rely on. A citation that does not say what the model claims is common enough to check every time.
Separate facts from inferences
Ask the model to mark each statement as sourced fact, inference, or estimate. Most confident-sounding paragraphs are two facts and five inferences.
Have a second model check the first
Paste the findings to another model with the sources: "What here is unsupported, outdated or contradicted?" Different models have different training and different blind spots.
Keep the evidence in the Project
Sources as links or uploads, the findings as a document, the open questions as a list. Six weeks later you can ask "what did we find about pricing in Germany?" and get the source, not a memory.
Verify before you act
Anything that decides money, a launch or a legal question gets opened, read and confirmed by you. AI narrows the search; it does not replace reading.
Three kinds of research
What to ask for, and what "done" looks like
| Research | Ask for | Done looks like |
|---|---|---|
| Market | Who has the problem, how many, what they pay today, how they buy, what changed recently. Bottom-up numbers with assumptions. | A one-page brief with every number traced to a source or marked as an estimate. |
| Competitor | Direct and indirect alternatives, pricing pages, target customer, most-praised feature, most-hated gap (from their worst reviews), positioning sentence. | A comparison spreadsheet and one paragraph on the gap nobody fills. |
| Product | How people solve this today step by step, what breaks, what they say in forums and reviews, what "good" would look like to them. | A list of jobs, pains and quotes, in the customers' words, ready for a spec. |
Reviews are the best source
For competitor and product research, two- and three-star reviews on app stores, G2, Capterra and Trustpilot are worth more than any homepage. Upload screenshots or paste the text into the Project and ask for the recurring complaints.
Comparing models
When two models disagree about a fact
For market research specifically, with Perplexity Sonar Pro doing the sourced sweep, see market research with AI. Send the same research question to two or three models at once. When they agree with sources, move on. When they disagree, you have found the thing to verify yourself, and usually the thing that matters. A disagreement about a competitor's price is a five-minute check of their pricing page. A disagreement about market size is a sign that the assumptions differ, and the assumptions are what you need to think about. Treat disagreement as a map of where the research is thin.
Free is the start. The latest models are the difference.
Research is where model quality and sourcing show most.
The free plan is a fine way to learn the method. The paid plans add research with cited sources, independent verification passes for factual claims, and the latest models, which invent less and reason better about what the evidence means. If a decision depends on the research, that is the plan to do it on.
Newer models hold your whole Project in view instead of the last few messages, reason through trade-offs instead of picking the first plausible answer, and are wrong less often and with less confidence. For the questions on this page, that is the difference between advice that sounds right and advice you can act on.
Pro from $20 a month, cancel any time. The free plan stays free.
- The latest ChatGPT, Claude, Gemini, Grok and Perplexity models, all on the same Project
- Far more messages a day, so a working session does not stop halfway
- Modes where several models draft, challenge and verify each other's answers
- Sourced web research and independent verification for factual claims
- Larger uploads for reports, data and review screenshots
- Documents and spreadsheets generated from the findings
Start here
First messages for a research Project
State the question and what decision it feeds. Then start here.
MarketResearch the market for [product] for [customer]: size bottom-up, what they pay today, how they buy. Cite a source for every fact and mark estimates.
CompetitorsFind the direct and indirect alternatives to [product], including "do nothing". For each: price, target customer, most-praised feature, most-hated gap, with sources.
ReviewsHere are competitor reviews. What are the recurring complaints and praises? Quote them. What gap do they point to?
Fact checkHere are findings from another model with sources. What is unsupported, outdated or contradicted? Be specific.
Facts vs inferencesMark every sentence in this brief as sourced fact, inference or estimate. Rewrite the inferences as questions to verify.
BriefTurn the research in this Project into a one-page brief for a decision on [decision], with the open questions listed at the end.
Honest answers
What people ask
Can I trust AI research?
You can trust the method, not the first answer. Sources you open, a second model checking, and your own verification of anything that matters make AI research faster than doing it alone and about as reliable as you make it.
Which AI is best for research?
Models built to cite sources are the starting point; models strong at reasoning are the check. Using both is the point of doing it in one workspace.
Does MultipleChat search the web?
Yes, where research and web-search features are enabled on your plan, with sources you can open.
How do I research competitors' pricing accurately?
Ask for it with sources, then open every pricing page yourself and screenshot it into the Project with the date. Prices change; your screenshot is the record.
Can I upload my own reports and data?
Yes. PDFs, documents and spreadsheets go into the Project and become part of the research context. See analysing files with AI.
What is free?
Your first Project, file uploads and the major model families. Sourced research, verification passes and the latest models are on paid plans. See plans.
Ask the question. Ask for sources. Ask a second model to check.
Put the evidence in a Project so the research is still useful in six weeks. Free to start.
Also: analyse files with AI · validate a business idea · compare AI models for a project · start a business with AI