I've tried to use Zotero over the years, and I always stumble with it because it didn't have a good handling of arXiv id's. The arXiv is one of the primary way of finding papers in my field (cosmology), and I really want to support ingestion via them, looking up actual paper details from them (while retaining the arXiv id), and including them in bibtex exports. I check back from time to time, but I don't think that's massively improved, but I would be very happy if anyone else knows otherwise.
Although it's gotten a little maligned here, I've actually switched to Readcube Papers recently, it supports everything I list above, and has been pretty good for everything else. It's a closed source and subscription which isn't ideal, but in the end I'm willing to pay for something that supports my workflow. I do worry a little about getting locked in, but ultimately, provided I'm willing to ditch any annotations, I think I can just export everything as bibtex and import to something else.
Zotero has supported adding items via arXiv ID since 2018, and saving from arXiv.org since forever. There's not currently a dedicated field for arXiv IDs, but they're added to the Extra field and included in exports (and accessible however you want via Better BibTeX export).
Although it's gotten a little maligned here, I've actually switched to Readcube Papers recently, it supports everything I list above, and has been pretty good for everything else. It's a closed source and subscription which isn't ideal, but in the end I'm willing to pay for something that supports my workflow. I do worry a little about getting locked in, but ultimately, provided I'm willing to ditch any annotations, I think I can just export everything as bibtex and import to something else.