How to get an AI research report with real sources
A chat model answers from memory and sounds confident either way. A research run goes out, reads dozens of pages, and comes back with a report where every claim points to where it came from. The second is what you want for anything that matters: a market, a competitor, a regulation, a purchase.
What deep research does differently
It plans the question into sub-questions, searches each, reads the results rather than the snippets, discards weak sources, and writes a structured report with inline citations. The output is longer and slower than a chat reply, typically a few minutes, and it is checkable: you can open every source.
The limit is the question. 'Tell me about electric bikes' produces an encyclopedia entry. 'Compare the five best-selling mid-drive e-bikes under 3,000 euros in the EU in 2026 on motor, battery, warranty and service network, with sources' produces something you can act on.
Writing a research brief
State the decision the report supports, the scope (region, period, price range), the dimensions you want compared, and the output shape (table, ranked list, pros and cons). Ask for sources to be dated and for disagreements between sources to be flagged rather than smoothed over.
How to do it in Vyndexo
Open Research
Pick Research from the sidebar and paste the brief. Cost is shown before the run; it draws from Tools credits.
Let it run
The run searches, reads and drafts. You can keep working; it reports back when finished.
Check the sources
Every citation is a link. Open the ones behind the key claims before you rely on them.
Reuse or schedule
Save the brief as a Skill to rerun it, or schedule it to repeat weekly with Scheduled Tasks.
Common questions
How is this different from asking ChatGPT?
A chat answer comes from the model's training; a research run reads current web pages and cites them. For anything recent or verifiable, the research run is the reliable one.
Which model writes the report?
You choose. Research works with the chat models in the workspace, so the same brief can be run with GPT, Claude or Gemini and compared.
Can it read PDFs and my own files?
Yes. Attach files to the run or keep them in a Space; they are used alongside web sources.