My research team made the switch from a manual EndNote workflow to Elicit for literature review. The promise was better discovery and organization. On discovery, it delivers. On organization and integration with my actual writing workflow, it's a significant step back from my existing Zotero setup.
The core issue is data portability and hygiene. Elicit locks your annotated PDFs and notes inside its ecosystem. My Zotero plugin (with Better BibTeX) gives me:
* Clean, exportable BibTeX keys for seamless LaTeX integration.
* Local PDF storage I control, synced across devices.
* Direct citation in Word/LibreOffice without a middleman.
* Structured data fields that don't break when I move systems.
Elicit's AI summaries are useful for a first pass, but they become a liability when I need to trace an insight back to a specific page in the PDF for my own manuscript. The workflow feels brittle—great for asking questions of a corpus, but poor for building a lasting, citable reference library.
For anyone doing serious academic writing, the inability to cleanly export your work is a deal-breaker. I'm now using Elicit only for initial paper discovery, then immediately exporting metadata to Zotero for actual management. The value isn't in the replacement, but as a pricey front-end filter. Is anyone else running this hybrid model, or has Elicit's export improved in a way I've missed?
I'm a consultant who sets up research and publication workflows for academic labs and professional services firms, typically dealing with 10-50 person teams. My own production setup for client work is Zotero with the Better BibTeX and ZotFile plugins, synced to a self-hosted group library.
* **Data Sovereignty & Export:** Zotero wins outright. All your PDFs and metadata are stored locally or on your own WebDAV server. You can export your entire library to a portable, standards-compliant format in minutes. Elicit, in my experience, keeps your annotated documents in its cloud, and structured exports are limited. You're building a reference library in a rented room.
* **Writing Workflow Integration:** Zotero is engineered for it. The Better BibTeX plugin generates persistent citation keys (e.g., `smith2023impact`) that don't change, making LaTeX writing bulletproof. The Word/LibreOffice plugins insert citations directly from your local database. Elicit is disconnected from the drafting process; it's a discovery and reading tool, not a citation manager.
* **Total Cost of Ownership:** Zotero's core software is free; you pay for storage syncing (roughly $20-$60/year for personal use, more for groups). Elicit operates on a credit system; heavy users conducting many literature queries can burn through $10-$30 in credits per user monthly, which adds up quickly for a team.
* **Primary Use Case Fit:** Elicit excels at the initial "finding and filtering" stage. Its AI summaries for a batch of papers are useful for quick triage. Zotero has no equivalent AI function. However, Zotero is built for the long haul of a research project: organizing, annotating, and citing hundreds of sources over years without lock-in.
My pick is Zotero for anyone who needs to write formal papers, theses, or reports that require precise citations and lasting reference management. Use Elicit as a disposable search engine. If you're deciding, tell me the average number of citations in your team's final manuscripts and whether you use LaTeX or a word processor.
Integrate or die
You're absolutely correct on the data sovereignty and workflow integration points. The consultant perspective on "building in a rented room" is particularly apt.
Where I'd add nuance is on the **Total Cost of Ownership** for teams, which extends beyond subscription fees. Implementing and maintaining your described Zotero setup - with self-hosted WebDAV, plugin management, and consistent citation key protocols across a 50-person team - requires non-trivial internal technical oversight. That's a real cost, even if it's not a line item to Zotero. Elicit's model essentially bundles that systems administration into its fee.
For some labs, that's a worthwhile trade-off. But it does create the exact lock-in you describe, making exit costs high once a team's annotated literature base grows inside the platform. The financial TCO calculation must include that future migration risk.
That workflow you described - using Elicit for discovery but reverting to Zotero for the actual reference library - is the sweet spot I've landed on too. The AI summaries are a fantastic filter for the "maybe" pile, but they just don't stick the landing for serious writing.
You nailed the brittleness. I've had a similar issue where an interesting summary in Elicit points to a claim, but tracing it back to the exact text for a citation is a chore. It adds friction right when you need it least. My plugin setup, messy as it can be, at least gives me a direct path from my note to the highlighted PDF text in my own storage.
It feels like we're using two different classes of tool: one for thinking, and one for building.
Comparing tools one review at a time.