Is there an established policy for how to cite a paper you link to in a question/answer? If not, what does etiquette dictate appropriate for citations?

I used to be careful to include full citations at the end of a post as I would in an article, but lately I've stopped and begun simply linking to the paper with the author's name.

I recently realized that this might be perceived as lazy/rude, or perhaps hurt some citation metric if I'm not including the doi. What are the community policies/opinions on this matter?

  • 6
    $\begingroup$ It is definitely good to use the citation helper: meta.mathoverflow.net/questions/2583/… $\endgroup$ Commented May 26, 2022 at 15:58
  • $\begingroup$ @AndrásBátkai That is awesome, I agree with Asaf's comment -- if you wanted to post it as an answer I suspect the community would agree that it's a good standard. $\endgroup$
    – Alec Rhea
    Commented May 26, 2022 at 16:03
  • 5
    $\begingroup$ Whatever you do, please DO NOT link directly to a pdf. Link to an informative meta-page instead, such as an abstract with bibliographic information. Leave the decision to download the pdf to the reader. $\endgroup$ Commented May 28, 2022 at 7:10
  • $\begingroup$ @EmilJeřábek Good point, will do in the future and I’ll do my best to edit past posts and correct direct-to-pdf links. $\endgroup$
    – Alec Rhea
    Commented May 28, 2022 at 10:39
  • 3
    $\begingroup$ At a minimum, I think it good scholarly behaviour to include at minimum the authors and title, so that a human reader familiar with the literature can instantly recognise the reference, and if a link breaks, an alternative can be (almost always) found by a simple cut and paste of the given information in a search engine. $\endgroup$
    – David Roberts Mod
    Commented May 29, 2022 at 9:22
  • $\begingroup$ @EmilJeřábek What about slides or course notes, for which there may not be a meta-page? $\endgroup$ Commented Jun 3, 2022 at 22:51
  • 2
    $\begingroup$ @TimothyChow That’s unlikely, really: how would a link to the slides spring into existence in the first place, if it didn’t appear, say, on the author’s web page? Even Google cannot just pull it out of thin air. But anyway, if the original page looks unsuitable, you can at least include as much meta-information as possible in the text: don’t just write “see example.edu/~jnd/shhhj.pdf”, but “John Doe, On regular shmips, slides for a talk at the Blomp workshop, 2013, 65 slides, 874 kB”. $\endgroup$ Commented Jun 4, 2022 at 7:19
  • $\begingroup$ @EmilJeřábek What I have taken to doing in such cases (no DOI or arXiv link) is to link to the Wayback Machine rather than the author's web page. If it's not in the Wayback Machine then the Wayback Machine will usually offer the option of archiving on the spot. If it's a PDF file then I have only been archiving the PDF file. But it sounds like you're saying I should archive the webpage as well as the PDF and point to the archived webpage. I can see some value to that, although it would not have occurred to me to do this extra work until I saw your strong opinion about not linking to PDF files. $\endgroup$ Commented Jun 4, 2022 at 13:14
  • $\begingroup$ @EmilJeřábek See this MO answer for an example where linking to the Wayback Machine PDF file directly seems to be the natural thing to do. I admit that probably I should just edit the answer rather than add the link in the comment. Probably I'll do that after you see this comment. $\endgroup$ Commented Jun 4, 2022 at 13:20

2 Answers 2


I would say that to some extent this could be simply based on common sense. And perhaps you could take into account your estimate whether you think that your answer (and the link/citation) might be useful in the future.

Of course, the link is working at the time when you're posting it. But maybe the link stops working at some point. It would be nice if the reader would still be able to somehow identify the paper. (At least if you assume that somebody might want to look at your answer a few years later.)

For example, probably we could assume that DOI or arXiv-identifier are relatively stable - and if they are mentioned, the reader should be able to get the information about the paper. But maybe for some other types of URLs, it might be quite difficult to get the paper just from the URL.

  • I guess springerlink.com is quite a well-known case. There are about 1000 dead links to springerlink.com networkwide and about 500 on MathOverflow. Many such links have already been fixed. If a post says nothing beyond "see equation (3.1) in this paper", then the reader will a have hard time finding out what the paper was once the link stops working. If the post contains some additional information, then even if the link stops working, the reader will be able to find the paper.
  • To use an example of your own post, here is an example of a link which no longer works: "A striking result referenced in the Shulman paper is due to Colin McLarty, establishing that the NF axiomatization of what a set is yields a ${\bf Set}$ that isn't Cartesian closed." From the context, perhaps the reader could eventually be able to find the paper in question. But I guess it is nice that somebody made an effort and replaced the broken link with a DOI-link. (In this case, there is a full citation of the paper near the end of the post - so there is actually no problem here. But finding the paper just from this paragraph - and from the broken link - would be a bit harder.)

(I would like to make clear that by including the above links I do not want to blame specific users. After all, who could have known that some springerlink or projecteuclid link might stop working. But I wanted to include at least some examples where something similar actually happened. There are probably much more egregious examples - I did not spend that much time searching. In both cases I linked above there was enough additional information in the question or in the answers to actually find the paper.)

Additionally, I will mention that adding link to Zbl can lead to link back to the MO post from zbMATH review to the MathOverflow post. Although I learned about this feature only recently, from this post on meta (and the answers posted there): Are there policies to respect in order for MathOverflow Q&As to be tracked back in relevant zbMATH Open reviews?

Related (at least tangentially):

A quote by Todd Trimble from this conversation: "I also think you make a good point about Math.SE being mindful of serving as a repository for the general audience out there. This probably is a weakness of MO, that it is not as useful as it could be."

Keeping possibility of link rot in mind is one of ways how to make a site like MathOverflow (or Mathematics Stack Exchange) useful.

  • $\begingroup$ I think these are great points, thank you for the input Martin. $\endgroup$
    – Alec Rhea
    Commented May 26, 2022 at 17:12
  • $\begingroup$ Project Euclid seems to go out of their way to hide any relevant DOI. Do you know a way to find it? $\endgroup$
    – LSpice
    Commented May 27, 2022 at 3:37
  • $\begingroup$ @LSpice This seems tangential to this question (and answer), so I left some response concerning Project Euclid and doi in the MO editors' lounge. (Dead links are discussed in the Boulevard, too. And there is a separate question about PE: Can (and should) Project Euclid links be replaced in bulk?) $\endgroup$ Commented May 27, 2022 at 5:00

I post my comment as an answer: it is definitely good to use the citation helper (see New feature - citation helper). It makes citation easy, and puts all the information needed into the post. Ideal for a lazy person like me.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .