I've been exploring SciSpace (formerly Typeset) for a few months now, primarily for its literature review and citation features. With their recent push into the "systematic review" workflow, I'm curious if anyone has taken it all the way through a full production project.
My initial tests show promise, especially for:
* The de-duplication and filtering tools, which seem robust.
* The AI-assisted screening, though I'm always cautious about accuracy rates.
* Having the PDF analyzer, notes, and data extraction fields in one interface.
However, I'm hitting some friction points that make me question if it's truly ready for a complex, multi-reviewer systematic review. My main concerns are:
* **Collaboration granularity:** Can you assign specific articles to specific reviewers, or is it a free-for-all? In my trial, it seems everyone can screen everything, which is a workflow killer for large projects.
* **Export flexibility:** How usable is the extracted data? Can you cleanly export to tools like Excel or statistical packages for analysis, or are you locked into their tables?
* **Protocol adherence & PRISMA:** Does it help enforce your pre-registered protocol during screening, or can reviewers easily deviate? And does it auto-generate a PRISMA flow diagram from your decisions?
I'm comparing it against tools like Rayyan and Covidence. SciSpace's integration of the writing suite is a big plus, but if the core systematic review management is clunky, it's a dealbreaker.
Has anyone here completed a full systematic review—from import to final write-up—using SciSpace? I'm particularly interested in:
* Real-world experiences with multi-reviewer conflict resolution within the tool.
* How you handled the data extraction phase and any schema limitations.
* Whether the final output (the written review) was easier to produce because everything was in one platform.
The collaboration model you described is exactly the bottleneck we encountered during a pilot. The lack of granular assignment forces you to manage workload outside the tool, using spreadsheets or manual tracking, which completely defeats the purpose of an integrated platform. For a project with three reviewers and 2000+ abstracts, this became a significant source of error and confusion.
On your point about export flexibility, the data extraction tables do export to CSV, but the schema is rigid and includes substantial internal metadata. You'll spend non trivial time trimming and transforming columns before analysis in R or Python. The lock in isn't absolute, but the friction adds overhead, especially when you need to re export multiple times during the screening process.
Protocol adherence is my biggest concern. The system doesn't actively gate decisions against your pre registered criteria; it's more of a passive reference. Reviewers can easily deviate without systemic checks, which is a critical flaw for maintaining methodological rigor in a production review.
--perf