I have a list of references in a pre-print article that I'm getting ready to submit to a peer-reviewed journal. I will need to change the format of the references according to the journal that I submit to. I have used citation generators in the past, but that takes a lot of time. I am wondering if people have tips on how to format references with ChatGPT. In addition, I'd appreciate advice on whether I'll need to disclose my use of ChatGPT, and if yes, whether that might lead to the manuscript being perceived less favorably.

  • 6
    Interesting. Many (most?) reference managers will reformat as needed, and they might solve a number of other issues now and into your journal future.
    – Jon Custer
    Oct 17, 2023 at 19:15
  • Did you manually format the citations initially? In my experience, if you have your citations stored in a reference manager or library switching formats is a non-issue.
    – sErISaNo
    Oct 18, 2023 at 6:35
  • At this point the references are in an ms word document in Vancouver format. I didn't use reference management software. I entered them manually. Thanks for the advice. Oct 19, 2023 at 1:34

3 Answers 3


In my experience using a reference manager, this is accomplished with one click or one command assuming the format is already available in the software; if not available, it's often downloadable in a template from the journal, and if even that isn't available then there are just a few more steps to create a format. With a long reference list it can take some time to process, but we're talking a couple seconds, nothing compared to the time it would take to check for errors that ChatGPT might introduce.

My advice would be not to do it, this is a task that is already solved well by multiple dedicated software packages; even among AI aficionados, this seems like a very low-opportunity area for the technology.

If you do choose to do it, you'll need to check that no errors were introduced in the process, which likely means checking each individual reference manually, and yes, you should disclose the use of ChatGPT for this purpose.


Assuming the references are currently in a format that cannot be processed by dedicated reference managing software, I think there are two ways that GPT could be helpful in getting them there:

a) Show GPT (some of) your references and ask it if it knows of some reference management software that can read them, and how to do it. It is possible that it knows a way that you don't.

b) Show GPT your references and ask it to write a piece of Python code that will parse that format and convert it to something a suitable reference manager can use.

Obviously, in both cases you should check that the results work. For pathway (a), if you have access to GPT-4, probably the version with web browsing will yield the best results. For (b), GPT-4 with the advanced data analysis plugin can likely do it directly (i.e. it can write the conversion code, run it in a sandbox, and give you the result), but I would expect you might have to iterate, depending on how easy your references are to parse mechanically.

With GPT-3.5, (a) is still worth a shot if you do not have any other ideas, since it will be easy to see if it suggests a tool and if that tool works.

In either case, your paper would not contain text generated by AI, so you should check the applicable policies of whatever journal you are submitting to, but I would think reasonable policies would not require such tool use to be disclosed any more than we disclose that Joe from Library Services at one point showed us how to use a reference manager.


I don't think this is something that has to be disclosed. Most people in academia use LaTex that automatically formats references for your depending on your commands (e.g. APA). I never seen anyone disclosing they are using LaTex to format references even if they do and I don't see chatGPT being different in this regard.

To do it just provide chatGPT with the information about paper X and then prompt it by saying "provide citation in APA style" or whatever style you want.

  • 5
    I guess the difference is that a number of journals and publishers have sweeping policies dictating that use of AI tools must be disclosed; see this set of examples.
    – Anyon
    Oct 17, 2023 at 23:08
  • 7
    Another difference is that ChatGPT does not necessarily generate accurate contents (it will seamlessly create fake content) so you would have to check and likely fix everything that it produces, which is.. counter-productive at best, likely to generate mistakes and misinterpretations at worst.
    – user126108
    Oct 18, 2023 at 1:30

You must log in to answer this question.

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