Deep Research agent: how we taught AI to do real, scoped research
The Deep Research agent is one of three AI agents in Contents Lab. It is the one that takes a research question and produces a brief, an outline, and a list of real sources.
What “real research” means here
A lot of “AI research” tools produce text that looks like research but is actually synthesis with no underlying sources. The Deep Research agent does the opposite — it produces a brief that is a list of findable sources, with citations you can verify.
How it works (high level)
- You give it a question (e.g. “What is the current state of the AI hardware market?”)
- It expands the question into a search plan (sub-questions, key terms, candidate sources).
- It runs the plan against real search and source APIs.
- It deduplicates and ranks the results.
- It writes a brief that summarizes findings, with inline citations to real sources.
- It saves the brief into your Source Library so it is one-tap-citable in any editor.
Why it does not hallucinate sources
Because every source the agent cites is fetched from a real, public search before it is cited. If the agent cannot find a source, it does not invent one — it says so in the brief.
Where it shines
The Deep Research agent is best for:
- Pre-writing research (before you draft a topic)
- Trend reports (weekly market scan)
- Source discovery (find 10 good sources for a topic you are planning)
It is not best for:
- Live fact-checking (use a dedicated fact-checker)
- Original interviews (it cannot call people)
- Internal documents (it works on public web sources)