The Research Agent helps you explore published research and synthesize knowledge — both from the broader scientific and academic literature and from the documents already in your hub. Ask it a research question, tell it what kind of answer you're looking for, and it will search, analyze, and deliver a fully cited report.
What the Research Agent can do
Search current, published research across open-access journals, publisher sites, and other legitimate sources — including up-to-date publications, not just historical material
Pull abstracts from paywalled articles and link you directly to the source so you can access the full text through your own subscriptions
Synthesize the contents of your hub so you can identify trends and takeaways without reading every document
Compare your hub's contents against the external literature to see how your internal knowledge stacks up
Cite everything in APA style, with direct links to each source
Running the Research Agent
Step 1: Enter your research question
Start with an open-form question. This can be broad ("What are the current research innovations in the consumer packaged goods industry?") or specific ("What are the five most published articles by Nature in 2024?").
Step 2: Choose your research question type(s)
Select the option that best matches what you're trying to learn. You can select more than one, but in most cases a single selection produces the most focused report.
Question type | What it does |
Landscape scan | Surveys what's currently published on your topic and summarizes the major findings and trends |
Evidence for a claim | Searches for published research that supports (or challenges) a specific claim — useful for validating product or marketing claims |
Specific paper lookup | Finds particular publications by author, journal, citation count, or other criteria |
Methodology | Focuses on research methods — how studies on your topic are designed and conducted |
Internal synthesis | Looks only at your hub and summarizes the knowledge inside it — no external search |
Hub vs. literature comparison | Compares what your hub's documents say against the published literature on the same topic |
Other | A flexible catch-all for research tasks that don't fit the categories above |
Tip: Selections act as focus areas, not filters. Selecting multiple types shapes what the report emphasizes rather than narrowing what it searches.
Step 3: Set your recency preference
Choose how recent you want the sources to be. Select a time horizon to limit results to newer publications, or choose No preference to search across all dates.
Step 4: Choose your synthesis depth
This controls how many sources the agent reviews and how long it runs:
Quick scan — A faster pass across fewer sources. Typically takes around 1-2 minutes.
Standard synthesis — Considers more sources while still factoring in time as a consideration. Typically takes around 2-3 minutes.
Deep dive — A thorough review across many more sources. Can take up to 5 minutes.
Please note: Because the Research Agent searches the live web, it runs longer than other Narratize agents. If your report is taking several minutes, it's working — not stuck.
Step 5: Add additional context (optional)
Use this field to add any caveats or specifics that should shape the research — for example, narrowing a methods question to a particular technique, or specifying an industry context.
Then run the agent. When it finishes, your report will be ready to review.
Understanding your report
Every report is structured around your question, but you can generally expect:
Major findings and takeaways, organized by theme or by source depending on your question type
Confidence notes explaining where the agent's confidence is high and where the evidence is thinner
Limitations, flagged transparently — for example, when an article was paywalled and only the abstract was available
A full reference list in APA style, with links to every source cited
How the Research Agent handles paywalled research
Much of the world's published research sits behind publisher paywalls. The Research Agent handles this transparently:
It looks for an open-access version first. If a freely available copy of a relevant paper exists, it will use that.
If a paper is paywalled, it pulls what's publicly available — typically the abstract — summarizes what it can, and clearly notes the limitation in your report.
It links you to the source. You can follow the link, log in with your own institutional or publisher credentials, and access the full text directly.
The Research Agent cannot log in to subscription services (such as SciFinder or publisher accounts) on your behalf.
Workaround for must-have papers: If a paywalled paper is central to your work, download it through your own access, add it to your hub, and run the Research Agent with Internal synthesis selected. The agent can then read and summarize the full text.
Frequently asked questions
Why didn't the agent analyze every document in my hub? When your question involves your hub, the agent reviews everything in it but pulls only the documents most relevant to your specific question into the report. The sources it used are always listed at the end.
Why is it slower than other agents? Live web search across many sources takes time. Quick scans usually finish in about 1-2 minutes; deep dives can run up to 5+ minutes.
Can it access my journal subscriptions? No. It will link you to paywalled sources so you can log in yourself, but it cannot enter credentials or access subscription content on your behalf.
Questions? Reach out to your Customer Success Manager or contact us through the in-app messenger.