Every claim can be checked
Source links sit next to factual claims and each statistic shows its date or period. If a source could not be opened, the brief does not pretend it was read.
DFIRST Deep research is an AI deep research tool for marketing questions. It searches the live web, gathers evidence, then writes a research brief with the question, key findings, conflicting evidence, limitations and source links. Each statistic carries its date, and evidence gaps are marked instead of filled with guesses.
Two shortened briefs. In the real result, every source is a link you can open.
Key findings
Conflicting evidence
People name health in surveys, while purchase data also points to price promotions.
Evidence gap
No dated figure for buyers aged 18 to 24, so it is marked unknown.
Key finding
Advice from pharmacists comes up again and again in consumer articles. Source: a 2026 pharmacy trade magazine.
Limitations
Most sources report what people say, not what they buy. Read as direction, not as size.
Three steps. The first one starts on this page.
Write what you want to understand in the box above. Name the market, the period and the kind of evidence you trust, for example surveys versus what people actually buy.
The first step searches the web and collects findings with their links and dates. The second reads that evidence and builds the brief from it, keeping every source.
You get a structured brief: the question, key findings, conflicting evidence, limitations, evidence gaps and sources, with findings kept apart from recommendations. It is saved to your session and your Data Room.
Source links sit next to factual claims and each statistic shows its date or period. If a source could not be opened, the brief does not pretend it was read.
The brief separates observations, estimates and gaps, and lists conflicting evidence instead of smoothing it over. When the evidence is thin, the brief says so.
The brief lands in your Data Room, where Firsty can read it in a session when you write the campaign, the brief or the deck that follows.
Evidence for a campaign idea or a positioning choice, with the sources to show the client.
A sourced answer to a market question before a meeting or a budget decision.
Background research for a pitch in a market or category the team knows less well.
The facts about Deep research, as it works in DFIRST today.
About Deep research and how it fits the rest of DFIRST.
Write your question in the box at the top of this page and continue in DFIRST. It searches the web for evidence, then writes a research brief with key findings, conflicting evidence, limitations and source links that you can open and check yourself.
Yes. Source links sit next to factual claims, the evidence ends with a list of sources, and each statistic carries its date or period. The tool is told never to invent a source or to suggest it read a page it could not open.
Treat the brief as a well-organised starting point, not a final verdict. It marks estimates and gaps, but web sources vary in quality, so open the links behind any number you plan to put in a deck or base a decision on.
No. It works from what a web search can reach, so reports behind a paywall or a login are not read. If a paid source matters, paste the relevant passages into your question and say where they come from.
The language of your question. Ask in Polish and the brief is written in Polish, while the source links stay as they are. Ask in English and you get the brief in English, whichever language the sources are in.
Your deep research draft waits on this device until your access starts. DFIRST opens to people on the waitlist first.
Updated