Say I'm reading a paper taken from my Zotero library which provides a really useful overview of a particular topic and I want to explore the papers it mentions throughout - is there an easy way to essentially copy and paste that paper's bibliography into my Zotero library? These won't necessarily be papers which I've already added to my library as they are papers which have merely been referenced by a paper I'm reading.

I hope this makes sense

The Reference plugin can the extract reference list from a PDF in your Zotero library. And can take you to the DOI/URL in your browser for any one that you do not already have. There of course you can choose to add/download it with the Zotero web connector.
https://github.com/MuiseDestiny/zotero-reference

Unfortunately the current documentation on github is out of date (only for the older-style, v6-compatible version). And the plugin's Settings dialog in Zotero is only in Chinese. But Chrome's inbuilt translation does a reasonable job on the github pages. And the rest is fairly intuitive (if necessary, Google Lens does a decent job of translating the Settings dialog). Someone has just added an Issue for English settings to be added (the older, v6-compatible version had that).
https://github.com/MuiseDestiny/zotero-reference/issues/355

But there's nothing in the Settings you really need to set in order to make it work (except for Chinese paper access via CNKI for those who need that). And it's so powerful that it is worth the effort to figure out how it works.

It can also show you a list of the papers that cite the current paper (with data from Semantic Scholar; for which a [free] API key is not required, but can be used if you get one). So the plugin's two display modes are References (that the paper cites) and Citations (that cite the paper).

Are you running Zotero v6 ? Because that looks like the Settings dialog for the old, v6-compatible version of the Reference plugin. The Settings dialog under Zotero v7 for the latest version of the plugin (1.4.4) looks like this, even on English language PCs:
https://s3.amazonaws.com/zotero.org/images/forums/u5906489/9icja67g6ss1v7l762ir.png

The old plugin version functions quite differently. It can still extract reference lists (but not get the cited-by list); but it does that in different ways. The new version is cleaner and more powerful (albeit with the only-Chinese Settings). On the plus side, the current github documentation (translated by Chrome for example) describes the old version. So that version may still be of use to you.

If you're OK modifying your workflow a bit, you can try a tool called RefTraceback available here:
https://sourceforge.net/projects/academic-tech-toolbox/

It's clunky but can save some time if you're researchin' hard.

Basically, it allows you to bulk-search the refs (you can extract the bibliography and select the refs that look interesting), and then you can save just the ones you need using the Zotero extension in you browser.

I've been meaning to develop this into a Zotero plugin, but have zero knowledge about how to do that. Any feedback/help welcome!

@ilia.leikin : Thanks for the link to RefTraceback!

Just in case it matters. I am on Linux, but I decompressed the zip file for RefTraceback, asked Gemini and Claude to turn it into something usable under Linux (pdftotext is readily available in most Linux distros) and, after one iteration, Claude (just plain Claude via web interface ---Claude Sonnet 4.5 Extended---, not Claude Code) produced a Python program that seems to closely match what yours does in Windows (with Gemini, what I used initially, I/"we" weren't able, after more than 10 iterations, to get a decently working thing). I can of course send you the code. I guess it should also run under Windows (not tested as I don't have Python installed in my Windows virtual machine).

As you say, it requires modifying workflow; but maybe pointing this out in the sourceforge page would be useful for people who are not on Windows?

Optimistic about the results, I asked for a simple minded Zotero plugin to do this from Zotero, something that would replicate what you do, but taking advantage of running in Zotero. But this did not work after many attempts (with none of Claude, Gemini, or ChatGPT). This could be just me, of course. I know some Python, the logic of your code seemed really clean, and building myself a mental model of what your code does after playing with it in a Windows virtual machine was not difficult, so giving clear instructions and making suggestions to Claude/Gemini was possible for me, but I do not know how a plugin works (nor any JS), so I was lost blindly following LLMs' non-working suggestions.

A couple of additions to this old/new discussion ...

The important settings in the Reference plugin are now in English. The github documentation remains related to the Zotero v6 version, which had a different interface and ways of working (but it's easy enough to figure out just by using it). The top section of the Settings in Chinese relates to the Chinese CNKI database.

There is a new AI-based plugin for extracting the reference list from a paper. I have only tried it with free Gemini and it kept throwing a limits-related error. YMMV. It does not try to do quite as much as the Reference plugin.
https://github.com/jmiba/Zotero-add-items-from-text

The Cita plugin was also trying to do reference extraction from early on in its development but I haven't looked at it recently, as its main aims were wider than that (sending data to Wikidata etc). I probably should look at it again as it looks to have matured a lot.
https://github.com/zotero-cita/zotero-cita?tab=readme-ov-file

Thanks both (@ilia.leikin, @tim820) for the additional details!

I tried https://github.com/zotero-cita/zotero-cita?tab=readme-ov-file a couple of few days ago. For the papers I've tested, it works great: it can query a range of external DBs and gets the reference right, which I think is a great way to get around the difficulties of parsing possibly complex documents. (I did not look in much detail to the rest of the extensive functionality). But, at least for me, it is not exactly what I am looking for, as it places the references as one or more notes, not as a clickable list of "in my zotero/not in my zotero, go grab it". Edit: this is only partially correct. Cita also puts the citations on a menu on the right, and you can also ask for it to "Auto link citations with Zotero items", so you can see which ones you have in your library and jump to them, or click on the magic wand for those not on your library to import them.

I didn't know about https://github.com/jmiba/Zotero-add-items-from-text . I just tried it briefly; I hit rate limits with free Gemini and could not get it work with DeepSeek, OpenAI, or Anthropic (timeouts or network errors). I did not try much, though, because it requires users to select the references and paste them in a box for processing.

In fact, playing around with Ilia's program (and unsuccessfully trying to vibe code a minimal plugin that would parse the references) made me realize more clearly what it is I am looking for and why (I am currently reading about fields that are new to me, and the "reference hoarding" is becoming a sticking point). I'd like to be able to mark/click on one or more references in a PDF (or epub or snapshot), find if they are in my Zotero library (if they are, be given the option of going there) or, if they are not, search for them in, say, arXiv, or OpenAlex, or Semantic Scholar.

Why? When I read papers I often already mark, in the pdf itself, what I want to check later. Having a separate list on the side of the PDF is, for me, cumbersome: I need to match the items in the PDF to the items on another list. Moreover, I might be interested in checking just a handful of references, but the full list might contain tens or hundreds.

The "Search selected text on google scholar" script from Actions and Tags (https://github.com/windingwind/zotero-actions-tags/discussions/535) is very neat and already does some of this: you can select text, and it searches in Google Scholar. I'll see if I can try to use Actions and Tags (with Claude's help) to build upon this to carry out the more elaborate procedure above and to search in arXiv, OpenAlex, Semantic Scholar, etc

Ideally, I'd like to just put the cursor in any line of a reference, and have code that would automagically select just the right amount of info (extending left and right as needed) for the search in my library and in external dbs to be as successful as possible. But, for now, that might be way too much. When playing with this yesterday, I found that correctly parsing references from possibly very different kinds of documents can be much more complicated than I naively expected, which probably explains efforts such as grobid (https://github.com/grobidOrg/grobid), Neural ParsCit (https://github.com/WING-NUS/Neural-ParsCit) and others (e.g., https://arxiv.org/abs/2205.14677 , https://arxiv.org/abs/2505.15948), as well as https://github.com/jmiba/Zotero-add-items-from-text .

And coming back to some of @ilia.leikin's comments,

a whole lot of references you've extracted from an article or a thesis you've been reading.
Because the moment you find an interesting reference and the moment you decide to add it to the Zotero database are distinct,

Cita, https://github.com/zotero-cita/zotero-cita, might already provide much of the code to do that: it leaves all the references as one or more notes, so one could either parse the notes or "Generate Report from Items" and parse the report. Getting all the references should be easy since most (all?) have a URL and a DOI.

Edit 2: Actually, Cita does a lot more! It can "Autolink citations with Zotero items" (on the Cita menu on the right, click on the three horizontal dots). So there we have it: the list of references, which ones are in Zotero, and for those that are not the option to import into the collection of choice!
(Is the zotero-actions-tags/discussions/568 script above therefore redundant? I do not think so. It can work for non-indexed references, including internal reports, student papers, etc. And it solves another issue ---at least for me---: getting to the reference while directly reading the possibly annotated PDF, instead of switching to the right to locate the reference. And Cita's reference list depends on upstream extraction quality ---Crossref, OpenAlex, and Semantic Scholar sometimes have errors or missing entries--- whereas the script always works from the actual text in the PDFBut certainly, I could make better use of Cita).

Edit: Related discussion in https://forums.zotero.org/discussion/126880/help-finding-a-plugin-for-importing-references-from-in-text-citations .

@ilia.leikin : the more I think about it, the more I think that Cita , https://github.com/zotero-cita/zotero-cita , already provides most of what I think you wanted. I'd like to make use of Cita's output for better extraction when on a specific PDF entry (I've written about it in "Extension: leverage Cita" in https://github.com/windingwind/zotero-actions-tags/discussions/568) but for what I understand you want, Cita I think would be the best. (Not perfect, though: I found sometimes it misses some citations, but that is a problem of the provider).

(My mistake for not having looked at Cita more closely earlier; it was not until @tim820 mentioned it that I got back to playing with it more carefully and even then, I was slow seeing several of the things it can do ---part of it might be related to the outdated documentation, so the best is to play with it and take a look at the issues, both open and closed, in github).