Elicit · Collaborative Sessions
Research with AI stops being a solo act.
Until last month, working with Elicit's AI meant working alone, with review happening afterward. The new Collaborative Sessions feature lets multiple researchers join the same session and work with AI on the analysis at the same time.
From "review it later"
to "work through it together"
Until last month, going back and forth with Elicit's AI was mostly a solo exercise.
Elicit added Collaborative Sessions, letting multiple researchers work with AI in the same session at once, according to its official site. Previously the assumption was a staggered workflow: one researcher would work with the AI to build a literature review or an analysis outline, then share it for others to review afterward. Collaborative Sessions is aimed squarely at removing that time lag.
Elicit is known as a research tool that uses AI to assist with literature review and evidence synthesis; the full product overview is on the Elicit product page. This addition reads as a shift in positioning — from an individual productivity tool toward something built into a team's actual research workflow.
| Before | After Collaborative Sessions |
|---|---|
| Working with AI was mostly a solo activity | Multiple people join the same session at once |
| Analysis results were checked in a later review step | Progress can be checked in real time, mid-analysis |
| Misalignment only surfaces when work is shared | Misalignment can be caught and fixed live |
Why research AI needed
"joining at once" at all
Most AI research assistants were designed starting from individual productivity.
In collaborative research, it's common for several people to be looking at the same body of papers, each querying the AI separately, then reconciling results afterward. Misreadings of the same question often don't surface until a shared meeting — by which point rework is already baked in. The value of Collaborative Sessions is catching that misalignment live, before it becomes rework.
This also reflects a broader move: an individual research AI tool absorbing a feature that used to belong to a separate category — team collaboration software. It's bringing the act of working with AI itself into an environment where co-editing tools like Google Docs or Notion are already the norm. The line between individual productivity tools and team collaboration tools is thinning, now in the research space too.
Who this actually affects, and how
Collaborative research teams
Teams doing multi-person literature reviews or systematic reviews get to align mid-analysis instead of after the fact — the more rework-prone the task, the bigger the win. Start by trialing it on one small joint review and comparing the efficiency against working solo.
Research PMs and managers
Makes multiple researchers' or analysts' analysis process more visible, potentially cutting down on individual status reports. Early on, plan out access permissions and session-management rules before rolling it out broadly.
Solo researchers
If your work is a self-contained individual project, this feature barely changes your experience. Your existing workflow holds up fine as-is.
Misalignment is cheaper to fix in the moment
than at the point you finally share it.
Worth thinking through before adopting
Real-time multi-user sessions are convenient, but there's a real risk that several people steering the same conversation thread can actually break the analysis's coherence. If it becomes unclear who asked which question and which conclusion was actually finalized, tracing back "why did we land here" later gets harder. The same "who decided what" problem that shows up in collaborative text editing can show up in a shared AI conversation too — worth watching for.
The practical near-term move is to trial Collaborative Sessions on a small joint review first, and settle your own rules for session record-keeping and division of responsibility before scaling it up. How well Elicit — a tool refined for individual use — holds up in reliability once team features are layered on is something that will only become clear with real-world use.