I'm starting to organize my literature review process and annotation is a big part of it. I've been using a mix of manual highlights in PDFs and Hypothesis for web-based papers, but now I'm evaluating SciSpace's built-in tool.
For those who have used more than one: what are the practical differences in daily use? I'm particularly interested in how well annotations sync across devices, and if you can effectively export your notes for use outside each platform. The vendor lock-in aspect worries me a bit.
Your worry about vendor lock-in is well-founded. SciSpace's annotations are convenient inside their ecosystem but notoriously difficult to extract in a usable, structured format. You'll get a CSV of highlights, but the context and connection to the source material is often lost.
Hypothesis, by contrast, uses an open standard. Your annotations are portable data. You can access them via API, and they remain anchored to the document via DOI or URL, not a proprietary file ID. For a literature review you intend to build upon for years, this is a critical architectural difference.
Manual PDF notes, while fragmented, are the only truly future-proof method. They survive platform demise. The real comparison is whether SciSpace's workflow benefits outweigh the risk of your notes being trapped in their platform. For a short-term project, maybe. For a PhD thesis or long-term research stream, I'd stick with Hypothesis for web content and a disciplined manual system for PDFs.
Show me the data
Hypothesis using an open standard sounds good until you have to deal with broken URL anchors or DOIs that get updated. Their API is fine for export, but good luck re-anchoring those notes to a local PDF copy you actually own. You're still trusting a service to remain online and maintain that link integrity.
Manual PDF notes "survive platform demise," but only if your notes are just highlights. If you're adding rich text notes in the margin, that's also proprietary data embedded in the PDF. Try extracting that cleanly into another format. You can't.
The real lock-in isn't the data format, it's the toolchain you build around it. If you can't script your annotation workflow end-to-end, including backup and search, you're already trapped.
Don't panic, have a rollback plan.
You're right about the toolchain being the critical layer. The exportable data format is only half the solution without a reproducible process to ingest and query it.
Regarding your point about broken URL anchors, this is where a dual-key approach helps. Hypothesis annotations can be anchored to both a public URL and a local file hash. If you maintain a personal archive of the PDFs you annotate, you can script a reconciliation step that matches the annotation's target source hash to your local copy when the URL becomes stale. It's an extra step, but it decouples the annotation from the service's link integrity.
The proprietary data issue with PDF margin notes is severe. While the PDF spec is open, each reader implements its own annotation format, creating a hidden lock-in. A scripted workflow must treat the PDF as a read-only source and store all rich text and metadata externally, in a simple, queryable store like SQLite.
—KM