Elicit's Research Agent is a powerful way to tackle complex research questions that go beyond what academic literature alone can answer. Unlike Elicit's Reports or Systematic Reviews, which are intentionally limited to Elicit's academic, journal, and clinical trial corpus, the Research Agent can pull from a wide range of sources (e.g. publications, public filings, press releases), produce flexible outputs, and let you iterate until you have exactly what you need. Once you have an answer, you can filter and export your sources, generate figures, and draft slides, all without leaving the conversation.
When to use the Research Agent
The Research Agent is the right choice when:
Your question requires sources beyond academic publications — like regulatory filings, press releases, product labels, etc.
You need a custom output format, such as a comparison table, competitive landscape, or executive summary
You want to explore a topic iteratively, following up and refining as you go — rather than running a single structured query
For questions that are well-served by academic or journal literature and a fixed report or systematic review structure (such as PRISMA), the Research Report or Systematic Review workflows are likely a better fit.
Choosing how much effort the agent spends
Every Research Agent session runs at an effort level, which you set from the slider in the prompt box before you send your question. The button shows the current level — click it to open the slider and move between Fastest, Fast, Balanced, Smart, and Smartest.
Effort controls how much work the agent does before it answers and how much of your monthly usage limit is consumed: how many sources it reads, how far it follows a lead, how much it checks its own reasoning. Moving toward Smartest produces a more thorough answer, takes longer, and uses more of your monthly usage. Moving toward Fastest gets you something back sooner and uses less usage. The popover shows the usage each level uses relative to Balanced.
Your choice applies to the session you're in. Every new session starts at Balanced again, so set the slider each time you want a different level.
Which level to pick:
Fastest and Fast: Scoping a topic, checking whether a question is worth pursuing, or asking something you already plan to follow up on.
Balanced: Most day-to-day research. Start here if you're not sure.
Smart and Smartest: Work you'll defend to someone else: comparing therapeutic approaches, sizing a competitive landscape, anything feeding a recommendation. Smartest searches most widely and reasons longest, and can take considerably longer to finish.
Effort levels draw from the same monthly usage pool as the rest of Elicit. See Usage limits in Elicit.
How it works
1. Enter your research question
Ask your research question in plain language as if you're talking to a human. If there is additional context you want to provide to help the agent get the answer you need, feel free to include it. The more specific you are, the better the agent will be able to target the data you're looking for. Before you send your question, you can also set how much effort the agent should spend on it — see the Choosing how much effort the agent spends section above.
2. Answer clarifying questions
Before the agent begins, it may ask a few questions to make sure it understands your needs — which sources to prioritize, how the output should be structured, and the scope of your question. Taking a moment here leads to much better results.
3. Watch the agent work
The Research Agent breaks your prompt into a systematic plan, then executes it step by step. You can watch which sources it's consulting and how it's building toward your output in real time. You're also free to leave the page and come back later when it's finished. Depending on the complexity of your request, it may take a little time for the agent to complete its work.
4. Review your output
All claims in the output are grounded in evidence, with citations you can trace back to the source. Unlike other AI tools, the Research Agent doesn't simply search and summarize — it builds a structured plan from your question and executes it reliably. Click on cited sources to learn more.
5. Iterate
This is where the Research Agent really shines. Once you have an initial output, you can:
Ask follow-up questions about the results
Request changes or additions to the existing output
Kick off entirely new analyses based on what you've learned
Ask the agent to generate different exportable artifacts — such as a report, tables, figures, slides, or compare/contrasts of different sources. Any artifacts generated will be available in the Artifacts dropdown menu.
A research agent session is a workspace for ongoing research, not a single query.
Working with your results
Once the Research Agent has an answer, you can turn that evidence into artifacts you can share and use — without leaving the conversation.
Filter, save, and export citations
When the Research Agent pulls papers into a table, you can act on those sources directly:
Filter — Apply rich filters to narrow down the set of papers
Save — Add papers directly to your Elicit Library
Export — Download the table as .bib or .ris citations to load into your reference manager
Generate figures alongside your analysis (paid plans)
Ask the Research Agent to visualize what it found — a bar chart of sample sizes, a timeline of publications, a comparison of effect sizes across studies — and it will produce the chart as a separate artifact next to the conversation, ready to drop into your own documents.
Draft slides to PowerPoint (preview — paid plans)
When you need to share what you found, the Research Agent can create basic slides for you. Just ask the Research Agent to create a slide deck for you. Download the result as a PowerPoint file and edit from there — a grounded starting point for presenting your research.
Note: Slides is a preview feature — it's shipping early so we can learn how you actually use it. If you try it, share your feedback at [email protected].
Sources the Research Agent can access
In addition to Elicit's corpus of 138M+ academic papers, the Research Agent can search:
The broader web
ClinicalTrials.gov and related clinical databases
Regulatory documents (e.g., FDA and EMA filings)
Press releases and company announcements
Product labels
Up to 20 files that you upload directly or select from your Elicit Library
Tips for best results
Be specific in your prompt. The more context you give upfront — focus area, geography, timeframe, output format — the better the agent can tailor its work. Use declarative language: give the agent tasks to perform or ask research questions; don't ask the agent for opinions or advice. Note that Elicit is intended to assist with research, data extraction, and analysis only. It is not intended to assist with writing or editing your own content.
Use the clarifying questions. Don't skip past them. These help the agent scope the task correctly before it spends time searching.
Iterate freely. If the first output isn't quite right, just ask. The agent can refine, expand, or reframe without starting from scratch.
Match the effort level to the stakes. Balanced handles most work. Move the slider up when the answer has to hold up to scrutiny, and down when you're scoping. Each new session starts at Balanced, so set it again when it matters.
Agent sessions are subject to your plan's monthly usage limit.
What is the Agent NOT good for?
The Research Agent will not provide support on how to use Elicit — it is designed only for research. For help using Elicit, please contact support at [email protected].
The Agent is not designed to write your content for you. It's designed to assist with research.