Extended Research and Go Deeper
Standard Deep Research is one agent investigating your question end to end. It is fast, it cites its sources, and for most questions it is the right tool.
Extended research is for the other kind of question — the one where being wrong is expensive. A board paper. A vendor decision. A regulatory position. Turn it on and Scholarly puts a team on the question instead of a single researcher.
What "a team" Actually Means
Extended research runs several specialists, and you can watch each of them work:
- Researchers — several at once, each taking a different part of the question, so breadth does not cost you the whole afternoon.
- Currency scout — checks what has changed in the last year, so a well-cited but outdated answer does not slip through.
- Fact checker — independently verifies the claims the answer rests on. It does not just agree with the researchers; it checks them, and it reports back Confirmed, Corrected, Contested, or Unsupported for each claim.
- Reviewer — critiques the draft before it is finished, the way a careful colleague would.
- Analyst — builds the calculations and charts, when the question needs numbers rather than prose.
They message each other while they work. The progress view shows who did each step and what they said to whom, so the reasoning is legible rather than a black box.
Turning It On
In the research composer, open Options and switch on Extended.
Because it is expensive to run, Scholarly confirms before it starts, and tells you plainly what you are committing to:
- Uses 3 AI creations (standard research uses 1).
- Takes about 10–20 minutes.
- Runs parallel researchers, a fact checker, and a reviewer.
Extended research is available on paid plans. On Free, the toggle shows an upgrade path instead. See Plans and Limits.
You can also just ask for it in AI Chat — "run extended research on this" — and the same confirmation appears.
Go Deeper
Go deeper runs Extended research on top of a report you already have.
Open any finished report and press Go deeper. A research team picks up where that report left off and chases:
- Newer data than the first pass found.
- Primary sources, where the first report leaned on summaries.
- The open questions the report itself flagged.
- Anything the first report could not verify.
You get a new, complete report — titled Deeper:
Go deeper costs the same as Extended research: 3 AI creations. You can start it from chat too.
Following a Run
While Extended research is running, the progress view shows:
- The step it is on, and how long it has taken against the 10–20 min typical range.
- How many sources it has read and how many facts it has found.
- The latest sourced facts as they arrive.
- For Extended runs specifically, a running count of what the fact checker confirmed and corrected.
You can close the tab. The run continues and you are told when the report is ready. See Background Tasks.
Reading the Result
An Extended or Go deeper report opens the same way as any other research report, on three tabs:
- Summary — the answer, how confident it is, and what is still open. Tap an open question to ask it in the research chat.
- Sources — the sources the report actually cites, ranked by how often, with every retrieved source one tap away. Exports and citation lists share one menu, and citations copy straight to your clipboard.
- Report — the full formatted write-up.
Reports produced this way are labeled Extended or Go deeper in the report switcher, so you can tell at a glance which pass a conclusion came from.
When It Is Worth 3 Creations
Use it sparingly — that is the honest guidance, and it is what the product itself says. It earns its cost when:
- The decision is expensive or hard to reverse. Vendor selection, a hiring market, a compliance position.
- Currency matters. Anything where last year's answer is the wrong answer.
- You will be challenged on it. A fact-checked report with verdicts per claim is a different thing to defend than a summary.
- A standard report left real gaps. That is exactly what Go deeper is for.
For everything else — background reading, a first look at a topic, a quick comparison — standard research at 1 creation is the better trade.
Related
- Deep Research — the full research guide: sources, evidence filters, outputs, and jobs.
- Plans and Limits — AI Creation Credits and what is on each plan.
- Background Tasks — leaving a long run and being told when it lands.
- AI Chat — starting or deepening research from a conversation.
- Sources and Citations in Chat — checking the evidence behind an answer.