Zettelkasten Forum


Which sources go in your reference manager, and which don't?

Hello Zettlers! This is a slightly long post regarding a fairly straightforward question. I've done my best to organize and clarify my thoughts for your benefit, but feel free to skip right to my question:

Which sources go in your reference manager, and which don't?

Intro

I've been somewhat inactive here recently in part because I'm currently reading the new edition of @Sascha's book, Die Zettelkastenmethode (2025), which I'm finding very informative, helpful, and worth my focus. I have much praise with which to bestow the author whenever I finish reading. (My reading pace is very slow since I'm teaching myself German in order to read it, which is not something I'd recommend, but for me it is worth it not only to access this book before the English translation is published but also for other reasons.)

The book is very clear about how to build and use a Zettelkasten in a way that creates value at each stage of knowledge work (to me, this is Der Kern der Quelle). The book's clarity results in part from the author's own insights but also from the author's precision of language, which I find I can still access through the language barrier (hard work, yes; accessible, also yes).

Die äußere Form vs Die innere Form

Sascha differentiates between "Die äußere Form" and "Die innere Form":

"Die äußere Form. Sie betrifft alle Formalien, die ein Zettel erfüllen sollte. Ein Zettel sollte beispielsweise einen Titel haben."

"Die innere Form. Sie betrifft die Qualität, mit der die Formalien erfüllt sind. Ein Zettel sollte natürlich nicht nur einen Titel haben, sondern einen guten Titel." (Fast, 2025, p. 37)

Something that is not covered in depth is Die innere Form of references:

"Ihnen fällt nun auf, dass die beiden Bestandteile „ID" und „Referenzen" fehlen. Das liegt daran, dass beides Formalitäten sind, zu denen es nichts Substantielles aus Perspektive der Zettelkastenmethode zu sagen gibt. Technische Eigenheiten wie die automatische Generierung der ID durch Software oder die Zitierweise hängen von Ihren Entscheidungen ab und auch von sich ändernden Möglichkeiten der Software." (Fast, 2025, pp. 40-41)

While I understand why Sascha doesn't cover Die innere Form of references in the book, I'd like to discuss one aspect of it here: using a reference manager.

Why use a reference manager?

The questions of whether and how to use a reference manager must be answered with "it depends." Using sources and managing references are givens in knowledge work, but how you do so depends on the context of your knowledge work.

So why even use a reference manager in the first place?

@Andy made comments on a related post ("The reference manager for note-taking") which effectively address the question, "Why use a reference manager?":

"The reference manager is like an interface between my personal knowledge base (my set of personal notes) and the collective knowledge base. The reference manager contains my set of references, which is the subset of the collective knowledge base that I have found to be relevant to my personal knowledge base. My set of references constitute a system that is managed systematically in the reference manager, just as my set of notes constitute a system that is managed systematically in my note system."

"When I use my note system to write for publication, I will take material from my note system and put it in a separate document for publication. When I run Pandoc (with a citation processor such as the built-in CiteProc, or an auxiliary LaTeX package such as BibLaTeX, combined with a citation style file), the citation keys are converted to properly styled inline citations, and a properly styled reference list is generated at the end of the document or at the end of each chapter. This ability to delegate the tedious busywork of formatting citations and reference lists to an automatic program is another primary reason for using a reference manager."

From this, there are two benefits of using a reference manager in my case: (1) to act as a conceptual bridge between external sources and my personal knowledge base and (2) to automate the tedious process of generating properly formatted citation keys and reference lists.

However, maybe you never intend to publish any writing, and the extent to which you need to use your references ends at the boundaries of your Zettelkasten. In that case, you might not need an external reference manager. You could use a reference management system that's contained to your Zettelkasten — i.e., you generate cite-keys and full references for each note and manage citations all within your Zettelkasten.

Which sources go in your reference manager?

If you do use an external reference manager, there is still the question of which sources go in your reference manager and which don't.

Part of me thinks that every source I reference belongs in my reference manager. The ease of a simple reference management system where all references are managed by a single system is clear and straightforward to me. In this case, there is no question as to whether a source should or should not belong in your reference manager: if you are referencing an external source, you add the reference to the reference manager. You don't have to think twice about it. This approach also scales easily — i.e., as your Zettelkasten grows and your reference manager grows alongside it, there is still no doubt as to how you manage your references.

However, another part of me wonders if this is always practical or even desirable. For instance, sometimes I reference irregular sources. What I mean by "irregular sources" is sources that are minor, ephemeral, or otherwise for which there is not a structured way to input metadata or an established approach for referencing. For example: movies, music, videos, emails, conversations, webpages, social media or forum posts/comments, etc.

A good example I can give is when I want to reference a webpage that I think I'm only going to reference once in a single note. In this case, the reference is minor in the sense that it is only intended to be used once in a single note. Additionally, perhaps the knowledge contained in this note is only for me and will never be published. Consider also that a webpage is fairly ephemeral in the sense that the links are not permanent. The content of webpages unpredictably change or disappear entirely (although there are ways of archiving webpages to increase their permanency).

For the most part, there are established conventions for referencing difference types of webpages. But this is not always the case for other types of sources.

It might not make sense for such minor, ephemeral, or non-standard references to be added to an external reference management system. Perhaps these references are better handled informally, contained entirely within individual notes, because they don't need the overhead of formal metadata.

Summary

I understand that this is one of those "it depends" questions, which relates to why Sascha doesn't cover it in the book. But that's also why I want to hear from others here!

Which sources go in your reference manager, and which don't?

Final word: I don't think the question, "Which reference manager?" is particularly relevant here, except regarding the varying functionality of different reference managers at handling irregular sources; however, I will add for context that I've experimented with BibDesk in the past, which seems to work well for macOS and is used by many Zettlers here.


Fast, S. (2025). Die Zettelkastenmethode: Kontrolliere dein Wissen. Independently published.

Comments

  • @dylanjr

    Which sources go in your reference manager, and which don't?

    You must ask yourself why careful reference management is important.

    The risks of not having references of all types in a reference manager are future regrets. I once spent hours searching for a citation, Googling weird phrases, and trying to find the quote I recalled about young adults, blindness, and dating. Learn from my pain. Since we struggle to predict what we'll need, collecting citations protects the future you.

    You mentioned several irregular sources, and I'm only familiar with Zotero. Zotero gobbles up everything you mentioned, including artwork, maps, and statutes. With its browser extension, capturing webpages is trivial. Zotero captures a link and a "snapshot" of the HTML, protecting you from internet rot.

    If you plan to use any AI, they hallucinate references like drunk poets. Scary if you can’t fact-check them.

    Will Simpson
    My peak cognition is behind me. One day soon, I will read my last book, write my last note, eat my last meal, and kiss my sweetie for the last time.
    kestrelcreek.com

  • edited April 27

    I second @Will. The cost of deciding whether to store a material in your reference manager is actually substantial, more so if you include the case when you later regret not storing the material you find useful later. In some sense, I use it as if it's a data lake.

    I use Zotero, and even pay for the storage in order to keep most of text-oriented digital resources suitable as reference (PDFs for journal articles and books, epubs, web page snapshots, etc.). I also use it in place of browser bookmarks.

    By default, Zotero is ready to create an entry for the following item:


    Item Types

  • edited April 27

    @dylanjr: I don't have an explicit list of criteria for what goes in my reference manager, but one criterion is that the sources are all presumed to be of sufficiently high quality. That means that I must have checked a source sufficiently well that it seems to be a high-quality source according to my criteria. If I later discover that a source that I previously added has a serious problem, I remove it from the reference manager. (This could happen, for example, if an article is retracted. Zotero provides automatic alerts about retracted articles in one's Zotero library. BibDesk is my primary reference manager, but I use Zotero as an adjunct as I described elsewhere.) I have an informal mental blacklist of journals and publishers that I will never add to my reference manager because their quality standards are consistently poor or vary too much. There are many sources on the Web that I consult but don't put in my reference manager, such as discussions in this forum. I save interesting discussions from this forum in a folder on my computer, and I've added a few instances of hyperlinks to these forum discussions in my notes, but I don't put them in my reference manager. There are many potential sources in my web browser bookmarks and in other files; eventually some of them will be put in my reference manager when I have reviewed them. My reference manager is a library of high-quality sources that I have cited or want to cite.

    Separately, an important reason for using a reference manager that hasn't been mentioned is the portability and interoperability of bibliographic data. One advantage of using a standard file format and markup language like Markdown for our notes is that we can use the notes with different software apps and/or easily convert the notes to other standard formats. The same is true for our references: by using a reference manager, we are keeping our references in a data format that can be easily used by other programs and/or converted to other formats, including, obviously, different citation styles. By using a reference manager, you are keeping your reference data maximally portable and interoperable.

  • I agree - I put everything in Zotero rather than have to find the reference later if it becomes important to something I'm working on.

  • I think that not everything you read should be a citation. We need a selection criterion based on quality, interest, and reliability. Low-quality citations are landmines waiting to dismember your credibility.

    Forum posts can be wayfinders. When @Andy or others point out real gems, I snag the actual paper for Zotero and jot down the forum post link in my notes for that paper.

    Pro tip: If your source title includes three emojis or the phrase “ultimate truth,” consider reconsidering it.

    Will Simpson
    My peak cognition is behind me. One day soon, I will read my last book, write my last note, eat my last meal, and kiss my sweetie for the last time.
    kestrelcreek.com

  • edited April 28

    I've been influenced by the reliable sources guidelines on English Wikipedia, a helpful general-purpose discussion of the quality of sources; there are also separate guidelines for reliable sources in medicine. Sources that I put in my reference manager are analogous to what would be called reliable sources in English Wikipedia, but hand-picked by me.

    The perennial sources list on English Wikipedia is an interesting table of sources that Wikipedia editors consider to be whitelisted or blacklisted. As I mentioned above, I have an informal list like this in my head.

    However, if you are doing historical or ethnographic work, you will be using many more of what you call irregular sources.

    James Donovan (2023). "Apologia pro bibliotheca: more than information". Practical Academic Librarianship, 13(2), 3–7:

    Each library preserves a permanent record of the intellectual endeavors of its own users. Were it not true that libraries offer a privileged window upon the lives they serve, several ongoing projects would make little sense. Not only is the Library of Congress reassembling the original Jefferson books, but the free online catalog LibraryThing works to reconstruct the Legacy Libraries of significant historical figures like Susan B. Anthony and David Bowie. In each case it is not the naked information that provokes interest, or even the individual books, but rather their associations with one another, and with their users. People are rightly interested in not only the titles from the New York Society Library read by Alexander Hamilton and Aaron Burr, but the actual volumes held in their hands. Libraries reflect the stories of the people who chose and read the books.
    Post edited by Andy on
  • For web pages, I find myself adding a citation at the bottom of the the note, in the back matter, that contains author, title, publication date, and URL. I generate date-based, throw-away citekeys for these, essentially. This thread could be e.g. [#20250428refs]. I don't expect to use the references more than once, maybe twice, and I skip the BibDesk (or for most people: Zotero) overhead then.

    Articles go into BibDesk directly, no matter how short or long. Formalizing the metadata, DOI, attaching the occasional open access PDF -- all that works flawlessly in a reference manager and does provide long-term access benefits.

    I would probably add web links to my reference manager when it would create an archived copy of the web page automatically. That would sound like a benefit for long-term retrieval.

    Author at Zettelkasten.de • https://christiantietze.de/

  • Things I want to write about go in ref manager. Things I want to look up as reference usually do not- for these, I will save to laptop and use 2-3 keywords on my note to refer to them. I read tons of medical PDF articles, author's last name very helpful, so examples are "Johnson liver cyst", "Anderson adrenal", "Wilson thyroid".

    I also save URLs as shortcut files (so they can appear alongside other files when I search my laptop) and rename them and refer to them with 2-3 keywords, e.g. "Wiki separation of concerns".

    I think it would be too tedious to include every last source in a ref manager, especially those you have no intention to ever cite in formal writing.

  • @JasperMcFly, consider using one of the browser plugins for reference managers. They make saving 'things you want to look up as references' easy. There’s no need for sorting, renaming, or separate downloads of PDFs; the browser extension automatically grabs the PDF and files it with the citation. Persistent website snapshots are saved with a single click. Reference managers have advanced search and sorting capabilities, making finding the reference you want simple.

    Will Simpson
    My peak cognition is behind me. One day soon, I will read my last book, write my last note, eat my last meal, and kiss my sweetie for the last time.
    kestrelcreek.com

  • edited April 30

    Thanks @Will. I should probably give the PDF browser extensions another look. I do like my clumsy way of saving everything to my laptop so I can have single search for any source regardless of file type.

    Post edited by JasperMcFly on
  • Thank you all for sharing your helpful insights! It's been a busy week, but I've finally had a chance to catch up with this thread. A few things stood out to me.

    Zotero seems more popular and useful than I suspected. I'm not exactly sure what I didn't like about using Zotero as a dedicated reference manager — maybe I didn't give it a fair enough shot. Regardless, the ability to integrate Zotero with BibDesk, as @Andy suggested, seems useful.

    There seems to be two general approaches in terms of which sources go in one's reference manager and which don't:
    1. Everything goes in the reference manager.
    2. Only certain sources go in the reference manager.

    No one mentioned not using an external reference manager (something I brought up in my original post). I still think this is a valid approach for certain circumstances. It might sound cumbersome to manage references manually inside the ZK, but it's not difficult, and I think there are certain advantages.

    I really like @ctietze's approach to generating throw-away citekeys for references when skipping BibDesk. This approach could be easily adapted to manually managing references entirely inside the ZK.

    One of the best reasons to consider an all or nothing approach to using an external reference manager is something that @zettelsan said:

    "The cost of deciding whether to store a material in your reference manager is actually substantial"

    These kinds of decisions can take away from deeper work. That said, I'm not sure how practical it is to always store reference information on every source in an external reference manager. I'd say, "It depends." But I need to experiment more to find out what works in my case.

    I find the discussion of source quality really useful.

    I like what @Andy said:

    "I don't have an explicit list of criteria for what goes in my reference manager, but one criterion is that the sources are all presumed to be of sufficiently high quality.[…] My reference manager is a library of high-quality sources that I have cited or want to cite."

    In this sense, a reference manager is like a Zettelkasten: the quality of your reference manager is a reflection of the quality of your sources, just as the quality of your ZK is a reflection of the quality of your notes.

    The reference manager as a library connects to the article @Andy shared, "Apologia pro bibliotheca: more than information". If your reference manager is like a personal library of high-quality sources that you have cited or want to cite, and memory is the unique mission of a library, then it makes sense to carefully nurture the contents of your reference manager.

    Having a concept of blacklist vs whitelist sources would be appropriate and useful in this case. In my mind, I definitely have a kind of blacklist of sources that I avoid.

    @Will said:

    "Low-quality citations are landmines waiting to dismember your credibility."

    This is an excellent point of caution. As is your point regarding Ai:

    "If you plan to use any AI, they hallucinate references like drunk poets. Scary if you can’t fact-check them."

    I interact with Ai models all day in my work and personal life. It is true that while they are incredibly useful, powerful tools that many people are much too quick to dismiss or disregard, Ai models aren't a replacement for human thinking. Being able to fact-check responses — and, in a broader sense, critical thinking — is a crucial skill that will only become more relevant as Ai becomes more prevalent. I've especially noticed that the models I use are poor at producing properly formatted citations and prone to hallucinating bibliographic information.

    One final thing to reiterate is something @Andy said:

    "Separately, an important reason for using a reference manager that hasn't been mentioned is the portability and interoperability of bibliographic data."

    This is definitely an important reason for anyone considering using a reference manager.

    Thanks again for all the thoughtful responses! :)

  • @dylanjr said:

    No one mentioned not using an external reference manager (something I brought up in my original post). I still think this is a valid approach for certain circumstances. It might sound cumbersome to manage references manually inside the ZK, but it's not difficult, and I think there are certain advantages.

    Whether or not you use a reference manager is tied to how you want to use references in the future. I use Zotero for references related to my work. However, my Zettelkasten contains mostly non-work items and I don't find it necessary to use Zotero for that. I store any needed references within the body of a zettel, near the end. If it's a web page and I'm worried about that changing or disappearing, then I use the Internet Archive to reference it. But I do not use Zotero or any other formal reference manager in my ZK, the main reason being the overhead involved.

  • @GeoEng51 said:
    Whether or not you use a reference manager is tied to how you want to use references in the future. I use Zotero for references related to my work. However, my Zettelkasten contains mostly non-work items and I don't find it necessary to use Zotero for that. I store any needed references within the body of a zettel, near the end. If it's a web page and I'm worried about that changing or disappearing, then I use the Internet Archive to reference it. But I do not use Zotero or any other formal reference manager in my ZK, the main reason being the overhead involved.

    Thank you for sharing your approach! Now we have each approach represented in this discussion :)

    Avoiding unnecessary overhead is a good reason in favor of not using a dedicated reference manager alongside a ZK.

    Of course, you should do whatever works best for you.

  • @dylanjr said:

    The reference manager as a library connects to the article @Andy shared, "Apologia pro bibliotheca: more than information". If your reference manager is like a personal library of high-quality sources that you have cited or want to cite, and memory is the unique mission of a library, then it makes sense to carefully nurture the contents of your reference manager.

    Thanks, this is a nice statement of exactly what I intended to imply by quoting that article and not adding any commentary! I'm glad to see that the implication of the quote was as evident as I thought it was.

    @GeoEng51 probably doesn't cite many sources in his ZK, so it makes sense for him to avoid the unnecessary overhead of a reference manager. But if you anticipate that the quantity of your sources will eventually literally be the size of a small library, with thousands to tens of thousands of sources as in my case, I think you'll find that the "scalability" (as software people say) of a separate database for your library of sources is superior to keeping source data in the ZK.

    One of the best reasons to consider an all or nothing approach to using an external reference manager is something that @zettelsan said:

    "The cost of deciding whether to store a material in your reference manager is actually substantial"

    These kinds of decisions can take away from deeper work. That said, I'm not sure how practical it is to always store reference information on every source in an external reference manager. I'd say, "It depends." But I need to experiment more to find out what works in my case.

    I would say that you can't avoid making a decision about whether to add a source to your reference manager if you're concerned about the quality of sources, as I am. You lower the cost of deciding by having decision rules. My decision rules are mostly intuitive after years of use, but I could list some of them as I did above, and the rest could probably be listed if I did a cognitive task analysis with a think-aloud protocol.

  • @dylanjr said:
    One of the best reasons to consider an all or nothing approach to using an external reference manager is something that @zettelsan said:

    "The cost of deciding whether to store a material in your reference manager is actually substantial"

    These kinds of decisions can take away from deeper work. That said, I'm not sure how practical it is to always store reference information on every source in an external reference manager. I'd say, "It depends." But I need to experiment more to find out what works in my case.

    To avoid being perceived as someone with poor quality control, I’d clarify a potential disconnect regarding what “reference manager” means here, particularly in the context of deciding whether to include a reference.

    I tend to add resources to Zotero liberally because I use it as cloud storage, eliminating the need to store reference materials separately. Zotero’s ability to store notes and highlights alongside references makes this approach especially convenient.

    The connection between Zotero and my Zettelkasten is minimal. Zotero generates unique citation keys (via the Better BibTeX plugin), but whether a reference makes it into my Zettelkasten varies. It might not be used at all, or it may serve as a unique ID when I create a literature note for that reference. In my workflow, the quality filter operates at this boundary between Zotero and Zettelkasten.

    This setup allows me to add an academic paper to Zotero without hesitation, even if I haven’t fully read it but it was mentioned in a significant context. I can’t assess a paper’s quality without investing time to read it thoroughly, but storing it ensures I can revisit it later if needed.

    For example, I recently regretted not adding a reference to Zotero. I recall a podcast episode where a guest noted that death row inmates, when given a chance to express gratitude for the last time, almost always thank their mother but rarely their father. This intrigued me, but I didn’t create a Zotero entry because the episode otherwise seemed unremarkable. Now, in a new context, I want to revisit the guest’s identity or the study behind the claim, but I can’t because I deemed the reference’s quality too low at the time. This illustrates the risk of applying too strict a quality filter early in the workflow.

  • edited May 3

    @zettelsan said:

    I can’t assess a paper’s quality without investing time to read it thoroughly, but storing it ensures I can revisit it later if needed.

    For example, I recently regretted not adding a reference to Zotero. I recall a podcast episode where a guest noted that death row inmates, when given a chance to express gratitude for the last time, almost always thank their mother but rarely their father. This intrigued me, but I didn’t create a Zotero entry because the episode otherwise seemed unremarkable. Now, in a new context, I want to revisit the guest’s identity or the study behind the claim, but I can’t because I deemed the reference’s quality too low at the time. This illustrates the risk of applying too strict a quality filter early in the workflow.

    For me, the risk or issue that you describe in this scenario (which has happened to me many times: for example, when I'm listening to my local NPR radio station while doing other non-knowledge-related activities such as cooking meals) is not about the reference manager. This is an issue at the COLLECT/CAPTURE stage, to use the term from David Allen's Getting Things Done: it's about having a robust habit of capturing a fleeting note in some kind of INBOX whenever something catches your attention (i.e., what Todd Henry calls "stimuli") so that you have the option of turning it into a zettel and reference later. I have inboxes to address this issue, but that's not what my reference manager is for.

    So it's NOT that I'm "applying too strict a quality filter early in the workflow", rather it's that the reference manager comes at a different stage of the workflow for me than it does for you. Your reference manager plays a radically different role in your personal information workflow than mine does: you're using it as some kind of inbox and "data lake" (to use your term), whereas I have other things that I use for this function, and the reference manager only enters later in the workflow.

    These different information workflow architectures lead to radically different answers to the original question.

    Post edited by Andy on
  • @Andy, I'm glad I understood the implication of the quote :) I thought it was a very interesting article — nice share!

    Good point regarding source quantity and scalability. Also, I agree with you regarding your concern about the quality of sources.

    @zettelsan, thank you for clarifying. When I said, "I'm not sure how practical it is[…]," I just meant that I'm ignorant — I wasn't taking issue with the approach. But now that you've explained, I understand how the way you use your reference manager allows you to collect sources first, before deeply assessing source quality.

    The connection that Andy made to the concept of inboxes, a la David Allen, further clarifies this for me.

    @Andy said:

    "These different information workflow architectures lead to radically different answers to the original question."

  • @Andy said:
    So it's NOT that I'm "applying too strict a quality filter early in the workflow", rather it's that the reference manager comes at a different stage of the workflow for me than it does for you. Your reference manager plays a radically different role in your personal information workflow than mine does: you're using it as some kind of inbox and "data lake" (to use your term), whereas I have other things that I use for this function, and the reference manager only enters later in the workflow.

    This assessment is correct, as far as I see. I also do have my "inbox" for the "collect/capture" stage, and that is effectively an opportunity for me to decide whether to take the resource into Zettelkasten or not at all. Even if I decide to not make a reference to it at that point, I liberally err on saving the resource to Zotero. In some sense, that's the only difference here.

    It's a factor to consider, if you want your reference manager to double as a cloud storage for reading materials. I'm scatter-brained, and way too many times I regretted not keeping things around, though I also find myself saving quite a bit more materials than I ultimately need. It's a balance to explore.

  • While technically feasible, using the ZK as the sole source for references can make it harder to work with the result.

    Or maybe it doesn't, for you.

    I like an external reference manager because I can plug it into many other tools, including pandoc or LaTeX for long-form publications. If I only had the references I keep inside my notes, I would still be able to assemble texts. It's just a bit more manual.

    Author at Zettelkasten.de • https://christiantietze.de/

Sign In or Register to comment.