Answers from the live web, not training data
A competitor's price from their pricing page this week, a regulation as it stands today, a market report from this year. Chat models answer from what they read during training, which may be years old.
Market research with AI
Do market research with AI that cites its sources: Perplexity Sonar Pro, inside MultipleChat.
Ask an ordinary chat model about a market and you get a fluent paragraph from memory: a size, a growth rate, three competitors, some of it years out of date and none of it sourced. Market research needs the opposite: current facts, with links you can open, checked by a second model, kept where you can find them. That is what Perplexity Sonar Pro does inside a MultipleChat Project, alongside ChatGPT, Claude and Gemini for the reasoning. This page is the method.
Perplexity is included on the free plan; Sonar Pro, the higher-tier model, is on paid plans.
Why a search-native model
Sonar Pro looks things up. Chat models remember.
The difference matters more for market research than for almost any other task, because markets change and memory does not.
Every claim with a link you can open
Sonar Pro cites. You click, you read, you decide whether the source says what the model claims. That check is the whole difference between research and guessing.
Search for facts, reasoning for meaning
Sonar Pro finds the numbers. ChatGPT, Claude or Gemini in the same Project reason about what they imply, size the market bottom-up, and argue about the assumptions. Neither does the other's job well.
Same Project, same files, same question
In MultipleChat the search model and the reasoning models read the same Project. Ask Sonar Pro for the facts, then ask Claude to check them, without re-briefing anyone.
Web research with any AI
Every model in MultipleChat can research the web. Sonar Pro is the one built for it.
Web search is not limited to Perplexity here. Where web search is enabled on your plan, ChatGPT, Claude, Gemini and Grok can all look things up from the same Project, with sources you can open. That matters for market research because the models are good at different things, and you can send the same question to more than one.
The search-native sweep
Built around retrieval: broad, current, every claim cited. Use it first, for the facts: market figures, competitor prices, recent changes, where customers gather.
Web-enabled reasoning
With web search on, the reasoning models can fetch a source and interpret it in the same breath: what a regulation change means for a new entrant, whether a market report's method is credible, what a competitor's pricing page implies about their customer.
Same question, two searches, one Project
Ask Sonar Pro and a web-enabled reasoning model the same question. Where their sources agree, you are done. Where they differ, you have found the fact to open and read yourself. The Project keeps both answers and both sets of sources.
How to switch
In a Project, pick the model from the selector for each message. Send the sweep to Perplexity, then send "check these findings against their sources" to Claude or ChatGPT with web search on, without re-explaining anything; the Project already holds the context. The web search page covers what is available on each plan.
The method
Five steps from a question to a brief you can act on
Write the decision the research feeds
"Should I build an invoice-reminder tool for small agencies, and at what price?" Research without a decision produces a pile of facts. Put the decision in the Project first.
Sweep with Sonar Pro, asking for sources
Market size and how it was estimated, the customer segments, what they pay today, the competitors and their prices, recent changes in the market. Every answer with a link. Open the links that matter.
Mark facts, inferences and estimates
Ask the model to label each statement. Then have a reasoning model (Claude or ChatGPT, in the same Project) check the facts against the sources and flag anything unsupported or outdated.
Size it bottom-up
Ignore "the market is worth $X billion". Count: how many of these customers exist where you can sell, what share has the problem, what you would charge, what share you could reach. Two models, independently; where they differ by ten times, that assumption is the one to test.
Write the brief and keep the evidence
One page in the Project: the answer to the decision, the numbers with their sources, the open questions. Six weeks later you can ask "what did we find about pricing?" and get the source, not a memory.
What to research
Six questions, the Sonar Pro prompt, and the check
| Question | Ask Sonar Pro | Then check |
|---|---|---|
| Market size | "How many [customer type] are there in [region], and what sources estimate it? Cite each." | Open the sources; note the year and method. Rebuild the number bottom-up with a reasoning model. |
| Customer segments | "Who buys [category] today: segments, typical size, who decides, with sources." | Compare with your own interviews. Segments in reports are often broader than real buyers. |
| What they pay today | "What do [customers] currently spend on [problem]: tools, staff time, services? Cite pricing pages and surveys." | Screenshot pricing pages into the Project with the date. Prices move. |
| Competitors | "List direct and indirect alternatives to [product], with pricing, target customer and their most common complaints in reviews. Cite." | Read the two- and three-star reviews yourself. The complaints are your positioning. |
| Trends and regulation | "What has changed in [market] in the last 18 months: regulation, pricing, new entrants, exits? Cite each." | Ask a reasoning model what each change implies for a new entrant. Search finds; reasoning interprets. |
| Channels | "Where do [customers] discuss [problem] online, and which communities, newsletters and events reach them? Cite." | Visit five. Read the rules. This becomes the marketing plan. |
The rule that makes it research
Any number you will put in a plan, a pitch or a price gets its source opened and read by you. Sonar Pro makes that possible; it does not make it automatic.
Worked example
Invoice reminders for small agencies, in an afternoon
Decision: build it, and at what price? Sonar Pro sweep: number of agencies with 2 to 20 staff in the target countries, with the statistics office and industry-body sources; what accounting tools charge for reminder features, with pricing pages; the three tools agencies mention most in forums, with the complaint "too complex for a small team" appearing repeatedly. Check: Claude, in the same Project, flags that one market figure counts freelancers as agencies and that a competitor's price is from an old page; both corrected. Bottom-up: 50,000 agencies × 30% with late-payment pain × $29 a month × 1% reached in year one ≈ 150 customers and about $4,300 a month. Small, honest, and enough to decide. Brief: one page, every number linked, two open questions for customer interviews. The whole thing stays in the Project for the plan, the pitch and the launch.
Sonar and Sonar Pro in MultipleChat
What you get on each plan
| Free plan | Paid plans | |
|---|---|---|
| Perplexity model | Sonar, the standard search model, to try the workflow | Sonar Pro, the higher-tier Perplexity model, for deeper research with sources |
| Other models on the same Project | The major families to try cross-checking | The latest ChatGPT, Claude, Gemini and Grok models for the reasoning and the check |
| Messages | A small daily allowance | Far more, so a research session does not stop halfway |
| Files | A starter upload allowance | Larger uploads: reports, exports, review screenshots |
| Verification | Compare basics | Modes where models challenge and verify each other, and independent verification passes for factual claims |
Free is the start. The latest models are the difference.
Market research is the case for Sonar Pro.
The free plan includes Perplexity Sonar, which is enough to see how sourced research works. The paid plans switch it to Sonar Pro, the higher-tier model, and put the latest ChatGPT, Claude and Gemini models next to it on the same Project for the checking and the bottom-up sizing. If a price, a plan or a pitch depends on the research, that is the setup 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.
- Your first Project and file uploads
- The major AI families to try the workflow
- A small daily message allowance
- 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
- Perplexity Sonar Pro for sourced, current research
- Independent verification passes for factual claims
- Spreadsheets and documents generated from the findings
Start here
First messages for a market-research Project
Write the decision the research feeds. Then ask Sonar Pro, and then ask a reasoning model to check.
Sweep (Sonar Pro)Research the market for [product] for [customer] in [region]: size and how it is estimated, segments, what they pay today, competitors with prices, changes in the last 18 months. Cite a source for every fact.
Competitors (Sonar Pro)List the direct and indirect alternatives to [product], including "do nothing". For each: price, target customer, most-praised feature, most common complaint in reviews. Cite.
LabelMark every sentence in these findings as sourced fact, inference or estimate.
Check (Claude or ChatGPT)Here are findings with sources from a search model. What is unsupported, outdated or contradicted by the sources? Be specific.
SizeSize this market bottom-up from the facts in this Project. Every assumption on its own line. Tell me which assumption you are least sure of.
BriefWrite the one-page market brief for the decision "[decision]": the answer, the numbers with sources, the open questions to test with customers.
Honest answers
What people ask
What is Perplexity Sonar Pro?
Perplexity's higher-tier search-native model. It answers from live web results and cites its sources, which makes it suited to research where currency and verifiability matter. In MultipleChat it is one of the models available in a Project, alongside ChatGPT, Claude, Gemini and Grok.
Is Perplexity on the free plan?
Yes, the standard Sonar model, so you can try sourced research. Sonar Pro, the higher-tier model, is on the paid plans. See plans.
Can I trust the market sizes it gives me?
Trust the sources it gives you, once you have opened them. Then rebuild the number bottom-up with a reasoning model. Top-down market figures from reports are a starting point, not an answer.
Why use ChatGPT or Claude as well?
Search models find facts; reasoning models interpret them, size markets from assumptions and argue about what the evidence means. In one Project you can do both without re-briefing.
Can I add my own reports and data?
Yes. Upload PDFs, spreadsheets and review exports to the Project; every model can then use them together with the web research. See analysing files with AI.
Is my research private?
Your Project, files and conversations stay in your account and are never used to train any model, by MultipleChat or by the AI providers, Perplexity included. See trust.
Write the decision. Ask Sonar Pro for the facts. Ask Claude to check them.
One Project, one afternoon, a brief with every number linked. Free to start.
Also: research with AI · validate a business idea · compare AI models for projects · start a business with AI