Wikipedia talk:Citing sources

In the middle of the preceding very long discussion about WP:CITEVAR the question of the value of bibcodes was raised. That topic got mixed up in the other one. So I want to open the question about bibcodes cleanly. Here is my proposal:

• Bibcode values should be added to a citation only when the linked page provide unique information unavailable through other links in the citation.

In my experience DOI has effectively replaced bibcode. The bibcode typically only shows the same info as the citation. I've compiled some examples here. The only exception I've seen is some older astronomy sources which are archived at the bibcode site and in these cases a bibcode would be fine. The proposal would apply to bots as well. Johnjbarton (talk) 23:16, 10 December 2025 (UTC)

Support (Obviously, but I want to encourage other editors to be clear with respect to the proposal). Johnjbarton (talk) 23:45, 10 December 2025 (UTC)

• First a disclaimer: I have no idea what a bibcode is… but based on the above discussion… if a bibcode simply repeats information already in the citation, I see no reason to include it. Exceptions can be made for specific citations where the bibcode does provide additional information (like a link to an abstract). That would probably mean adding it by hand, rather than by bot. Blueboar (talk) 01:21, 11 December 2025 (UTC)

Support Please yes. I waste a not insignificant amount of time checking these junk identifiers added by citation bot and removing them. Went through the same thing with s2cid back in the day. Esculenta (talk) 01:04, 11 December 2025 (UTC)

Support. Makes sense. Additional information (metadata) is in my opinion usually helpful and should not just be forbidden because "some editors don't like it", but repeating information already accessible via other existing metadata (such as the DOI) adds nothing of use. Gawaon (talk) 02:44, 11 December 2025 (UTC)

I've no opinion about the value of Bibcode, as I don't use it myself – it is, after all, an "identifier used by several astronomical data systems", so only people editing astronomy-related articles are likely to see it – but the proposal creates some fortune-telling problems for editors:

• Alice adds a citation with a Bibcode in 2024.
• Bob adds a different identifier in 2025. The new identifier duplicates all the Bibcode content (and maybe adds something else).
• Now Alice's addition of the Bibcode should be reverted, because the "the linked page" no longer "provide[s] unique information unavailable through other links in the citation"? Or maybe we all just need to make a mental note that this now-duplicative Bibcode should stay, because when it was "added" there were no "other links in the citation"?

WhatamIdoing (talk) 03:04, 11 December 2025 (UTC)

so only people editing astronomy-related articles are likely to see it. One could argue that those are the people most likely to see useful bibcodes, but because it has been incorporated into Citation bot's editing, bibcodes appear widely outside of the astronomy context, in everything from Abortion to Zoonosis.

In answer to your question, it would make sense to remove the bibcode with or after Bob's edit. Nikkimaria (talk) 03:09, 11 December 2025 (UTC)

Exactly. I expect astronomy articles will continue use bibcode, but not elsewhere. Johnjbarton (talk) 03:20, 11 December 2025 (UTC)

Would it be possible to set the bot up so it only triggers adding bibcodes when on astronomy articles (defined broadly/loosely)? Or alternatively only on citations to a broad whitelist of known astronomical publications. That might avoid a lot of the lower value ones, while keeping them for the contexts in which they are ore straightforwardly useful. Andrew Gray (talk) 10:13, 14 December 2025 (UTC)

The proposal as worded doesn't apply to this case, where the bibcode was there first. It's only about adding bibcodes when all the supplied information is already there. And nobody has suggested a mass removal of existing bibcodes. Gawaon (talk) 03:30, 11 December 2025 (UTC)

If we adopt this rule, then there will be someone suggesting mass removal. WhatamIdoing (talk) 04:01, 11 December 2025 (UTC)

Well, should that happen, we can discuss it. Gawaon (talk) 06:02, 11 December 2025 (UTC)

• Support per Blueboar. Nikkimaria (talk) 03:10, 11 December 2025 (UTC)
• Support. More generally, I would support similar wording to WP:ELNO #1 for all reference identifiers (not just bibcodes): they should only be added when they provide a unique resource beyond what the reference itself already contains or should contain. So identifiers that provide article text (jstor, doi, most hdl, some bibcode) are generally ok, as are identifiers that provide reviews (mr, zbl) but metadata-only identifiers (most bibcode) are not. —David Eppstein (talk) 17:09, 11 December 2025 (UTC)

  That wording would ban ISBNs. WhatamIdoing (talk) 17:32, 11 December 2025 (UTC)

  I think not. ISBNs link through a Wikimedia page that, among other things, leads to Google Books pages for the book that might provide previews.

  It might ban ISSNs but I'm fine with that. —David Eppstein (talk) 18:39, 11 December 2025 (UTC)

  I only find ISSN useful in the case where a journal is obscure, especially if it has a common name and there is no publisher link or DOI for each paper. If the journal is called something like the Journal of Psychology but is the one published by a college in the Philippines instead of the one published by T&F, then throwing in the ISSN (2423-2084) could improve clarity and help a reader find the source (though disambiguating the journal title like Journal of Psychology (SDCA) or SDCA Journal of Psychology might also be a good idea). Adding the ISSNs to every citation seems unhelpful. –jacobolus (t) 19:21, 11 December 2025 (UTC)

  I also find ISSNs useful because it's more efficient to search Scopus with an ISSN than with a name. With an ISSN, you know that you're getting the right journal and not one that has almost the same name. WhatamIdoing (talk) 17:11, 13 December 2025 (UTC)

  An ISBN is a "reference identifier". If you restrict "all reference identifiers", then you restrict ISBNs. You can argue that Special:BookSources is "a unique resource", but others can argue that it's not, especially when it's not linked (example, example). They can also argue that a DOI is not "a unique resource" compared to a direct URL since, in the optimal case, they both end up at the same place.

  More generally, I think it's misguided because it assumes that there is only one valid way to use these "references identifiers". I can walk into a book store with an ISBN and order the book. Is that "a unique resource" within the meaning of WP:ELNO#EL1? I don't think so, but it's still a valid use of the reference identifier. Similarly, I suspect that just as I gravitate towards PMIDs instead of DOIs, other readers may have a preference for various identifiers (e.g., if they're using citation management software that plays well with Bibcodes but not with PMIDs). I don't see any reason to give them "my" preferred identifier and hide theirs. WhatamIdoing (talk) 17:18, 13 December 2025 (UTC)

Possible support with different wording. A bibcode links to a database, so by definition it gives more information than a doi alone. It is also being expanded to cover fields other than astronomy. The new SciX replacement coming has a neat field indicating whether the paper as been refereed. The question is what kind of information do we need for references? I propose including the bibcode only if it links to a full copy of the paper with no subscription required and the doi does not, or if the paper does not have a doi.

The bibcode and ADS have become so successful we may not need them for many astronomy articles anymore. The American astronomy journals have gone open access and also put their pre-1998 archives into ADS. A doi for one of these old astronomy articles is translated into a bibcode on access, and ends up at the database. See doi:10.1086/115657. StarryGrandma (talk) 20:52, 12 December 2025 (UTC)

If ADS hosts a full copy of the paper, I think it's fine to include a bibcode, irrespective of what the DOI has. If there is a DOI and a bibcode which lead to the same webpage, I think we should just use one of them (the bibcode is probably better in that case, but either would be fine). –jacobolus (t) 22:03, 12 December 2025 (UTC)

You have a good point. I just ran into a small journal where the full copy in ADS downloads easily but the journal site keeps timing out. StarryGrandma (talk) 23:09, 12 December 2025 (UTC)

In my opinion these cases are covered by the proposal. This is proposal is not against using bibcode when they are uniquely useful. It is against using them routinely even when they offer nothing but a waste of time. Johnjbarton (talk) 00:45, 13 December 2025 (UTC)

Why do say the bibcode would be better in such a case? DOIs are far more widespread, so they should be the better choice when in doubt. Gawaon (talk) 03:08, 13 December 2025 (UTC)

In many cases when full copies are available through multiple sources, they are behind different paywalls and different copies might be accessible to different readers. This often happens to me with articles on JSTOR (which I can read courtesy of my employer) for which the official doi version points elsewhere. The same is likely true with some ADS full-text versions. So I am happy to keep both bibcodes and dois for those ones; its the ones that don't provide full text that are problematic. —David Eppstein (talk) 03:22, 13 December 2025 (UTC)

But my reply was about the situation where both "lead to the same webpage", excluding scenarios such as the one you mention. It's more or less agreed that if the bibcode provides additional useful information (such as another full-text link), it's helpful to add and keep it. Gawaon (talk) 03:34, 13 December 2025 (UTC)

Let me be more precise:

• If there is a DOI for the publisher's website which points somewhere different than the bibcode, and the bibcode contains unique information (e.g. has a full-text scan, or some data files or code associated with the paper), then they should both be listed.
• If the only available DOI is just a redirect to the bibcode page, we should only keep one of the two; I think it's preferable to keep the bibcode in that case, because it gives readers a better idea where their click will go, but either one or the other would be fine.
• If there is a DOI or other link to a publisher's site or JSTOR or comparable and the bibcode for the paper leads to a page which does not contain unique information (e.g. only has basic citation metadata, or only citation metadata + abstract but the abstract is also available at the publisher's site) I think we should leave out the bibcode.

–jacobolus (t) 05:09, 13 December 2025 (UTC)

That seems largely equivalent to Johnjbarton's original proposal if one prefers the DOI when in doubt. And too complicated a solution won't work anyway, so that's probably for the best. Gawaon (talk) 09:02, 13 December 2025 (UTC)

"A bibcode links to a database, so by definition it gives more information than a doi alone." Please provide one example. I provided multiple examples of the opposite. Johnjbarton (talk) 00:43, 13 December 2025 (UTC)

The example doi:10.1086/115657 is registered by the American Astronomical Society itself, see https://api.crossref.org/works/10.1086/115657 . Which makes me wonder, is there any case where a bibcode leads to a fulltext which is not provided by the corresponding DOI's destination? Last time I checked, such cases were exceedingly rare (say, about 100 records out of 26 million ADS records), but right now I can't find a single one. There are some cases like records for 1970s articles in soviet journals which claim to have a PDF but actually don't. (Not sure if the translation allegedly attached to the record went missing in one of the recent pointless redesigns.) Nemo 14:03, 22 December 2025 (UTC)

I have seen cases where there was a scan in ADS and also a separate paywalled publisher link which did not provide free access to the full text (I don't have any examples at hand though). I have seen cases where there was a scan in ADS and a different scan freely available from the publisher. I have also seen cases where there was a scan available in ADS and as far as I know no DOI at all (e.g. Bibcode:1984QJRAS..25..126S). This proposal would support keeping bibcodes in all such examples. –jacobolus (t) 14:38, 22 December 2025 (UTC)

• Question, aren't DOI's subject to a lot of linkrot? This search shows 8,336 articles with broken-doi notes. Abductive (reasoning) 14:57, 13 December 2025 (UTC)

  Well, in theory they are meant to last forever. In practice, of course, things aren't always quite as neat as in theory. Gawaon (talk) 15:17, 13 December 2025 (UTC)

  As far as I can tell from

  • Singh, H., West, R., & Colavizza, G. (2021). Wikipedia citations: A comprehensive data set of citations with identifiers extracted from English Wikipedia. Quantitative Science Studies, 2(1), 1-19.

  in 2020 there were around 1 million DOIs in wikipedia, so that puts the link rot around 1%. In my experience that is low compared to other issues with citations. In any case, the proposal here is not affected one way or another. Johnjbarton (talk) 18:11, 13 December 2025 (UTC)

  Naw, I think that broken DOIs never get fixed and are removed from the citations over time. And also people shouldn't count on the PMID and PMC identifiers staying up, as they are US government websites currently controlled by anti-vaxxers. During the shutdown the US Census and GNIS websites went down. Abductive (reasoning) 21:25, 13 December 2025 (UTC)

  DOIs are still valid unique identifiers and remain so forever. If broken, they can be restored at a later date (I have seen that happen after someone inquired about one), and I wouldn't advice removing even a broken DOI from a reference. It's less useful than a working one, but not completely useless. Gawaon (talk) 02:43, 14 December 2025 (UTC)

  Some broken DOIs are fixable, they are simply edit errors. Examples: . I think the citation bot could repair these. Johnjbarton (talk) 03:30, 14 December 2025 (UTC)

  I have also encountered broken dois from still-active journals when they rearranged their web site. Contacting the publisher can often get them to fix these. —David Eppstein (talk) 08:37, 14 December 2025 (UTC)

  Yes, I've run into a few of those. In a few cases that can make the source inaccessible; it is better to have other alternatives, when possible. Praemonitus (talk) 05:50, 22 December 2025 (UTC)

  This remains irrelevant to the current proposal, which is not about alternative ways of accessing the source. When a DOI is broken, a bibcode which leads to an ADS page containing nothing but redundant article metadata and a "publisher link" consisting of a DOI is not going to be of any additional use to readers. If they click the bibcode and then click through to a broken DOI, they'll end up at exactly the same 404 page as if they clicked a broken DOI directly from Wikipedia. –jacobolus (t) 05:59, 22 December 2025 (UTC)

  What jacobolus said. But I'll answer the question anyway: it's pretty rare for DOIs to fail, although it happens more often with some incompetent legacy press. Some mistakes take longer than others to solve. A DOI will nearly always return at least the metadata record associated with it. There can be some complicated situations where it's not clear who "owns" a DOI (or not), e.g. if a publisher is failing but not completely dead yet; or when a publication is sold to a new publisher and remains in a limbo for a while. The full text is often not archived, but when it is the DOI eventually directs to an archive when the publisher fails. Nemo 14:03, 22 December 2025 (UTC)

• Oppose - I think that it is valuable for readers who look at citations (I grant that such readers may be in the minority) to have as many methods of getting to the source material as possible. This allows readers to exercise their own choice in terms of their preferred method of looking things up, and guards against future failure of other links. What harm does an additional link do? Cheers, SunloungerFrog (talk) 10:11, 14 December 2025 (UTC)

  We normally edit article to remove two sentences that say essentially the same thing, saving the one that provides the best explanation. We should apply that principle here, for the same reason. The vast major of bibcodes waste editor time when verifying sources. The ADS site is often slow and the page it delivers is almost always pointless. The bibcodes are extraneous junk characters except in few cases. By asking editors to only include bibcodes that provide value we make them much more valuable. Bibcodes are not an alternative lookup method as I demonstrated with my example above. Johnjbarton (talk) 16:45, 14 December 2025 (UTC)

  I think that there is a big difference between removing redundancy in article content, which I agree is a good idea, and removing redundancy of possible access to article sources, which I think is a bad idea. Cheers, SunloungerFrog (talk) 17:11, 14 December 2025 (UTC)

  The bibcodes under discussion here do not provide access to the sources. They only contain metadata and redundant links to the publisher's site. –jacobolus (t) 01:39, 15 December 2025 (UTC)

• Support: bibcode is the first or second worst offender in terms of cruft added to citation templates. It could be ok to add it to the wikitext if templates are modified to not display it. When the bibcode leads to a full text, the direct link to the PDF should preferably be in the URL parameter. Nemo 17:54, 14 December 2025 (UTC)
• Question Somewhere in the prior discussions about bibcodes, it was pointed out that the ADS website gives information on which articles cite which. It was argued that their citation graph is too incomplete for this to really be useful information. I'm not sure of that and am open to being convinced either way. (I think the relative completeness might be a field-by-field thing that probably depends on how much that corner of science uses the arXiv. Speaking very informally, the citation counts in ADS seem to be lower in my experience than those from Google Scholar, which isn't always a bad thing because GS picks up random PDFs from all over the place, and sometimes ADS catches things that GS misses.) Is there a consensus that the ADS citation graph is not useful information? Stepwise Continuous Dysfunction (talk) 21:37, 14 December 2025 (UTC)

  There are a large number of citation graph websites. I don't think linking (any or all of) them is useful enough to justify such links, especially since going to the site and typing in the DOI or other metadata takes only a few seconds for any reader who cares (or faster using a bookmarklet or browser extension), and then they can pick between Scopus, Web of Science, Google Scholar, Internet Archive Scholar, CiteSeer, OpenAlex, Semantic Scholar, ADS, OpenCitations Corpus, or whichever other, based on personal preference. Cross-linking all of the citation graphs seems like another thing that could be done at Wikidata or some similar venue, but it doesn't seem relevant to Wikipedia. –jacobolus (t) 01:48, 15 December 2025 (UTC)

  I updated my list of bibcode examples to include the number of citations reported by ADS and Google Scholar. Scholar, in my experience, is inflated a bit, including some sources that don't cite the root article but I guess the biggest issue is that the ADS database is limited and its citation tree is only among the articles in its database. Johnjbarton (talk) 16:45, 15 December 2025 (UTC)

  The ADS database is limited, and other citation-graph services exist, but does that make the citation information brought up by clicking a bibcode useless information? It's still information that no other link in the reference would immediately provide. Stepwise Continuous Dysfunction (talk) 22:54, 15 December 2025 (UTC)

  There has to be sufficient benefit to balance the extra distraction caused by adding each new identifier. I could plausibly imagine that for an article about astronomy or astrophysics, where the relevant part of the citation graph is mostly included in the database, many sources may have full text and/or extra data files available, and readers in the field may be heavy users of ADS, it might be worthwhile to include bibcodes for every citation which has one. But for Wikipedia articles about other topics, the content hosted at ADS is entirely redundant with the citation metadata already included in Wikipedia, the database only contains an arbitrary small fraction of the citation graph, and typical readers (including experts) are unfamiliar with bibcodes and destined to be disappointed or even confused if they click one. –jacobolus (t) 09:01, 16 December 2025 (UTC)

• Comment I could get on board with a rule that bibcodes generally shouldn't be included when arXiv IDs are, since the ADS page is only a click away from the abstract page that the arXiv serves up. Stepwise Continuous Dysfunction (talk) 22:55, 15 December 2025 (UTC)
• Oppose Bibcodes are useful. That you don't personally find them useful is not a reason to ban them. Headbomb {t · c · p · b} 23:22, 21 December 2025 (UTC)

  Please read the proposal: bibcodes are not being banned. Johnjbarton (talk) 23:33, 21 December 2025 (UTC)

  Although about a different context, WP:USEFUL is also relevant: to whom are these particular bibcodes (the ones we are discussing, the ones that do not provide information) useful, and why is that an encyclopedic purpose? —David Eppstein (talk) 00:26, 22 December 2025 (UTC)

• Comment: in many cases the ADS page linked by the bibcode provides a free ADS_SCAN, giving an accessible copy of the article. That's reason enough to retain bibcodes. But I'd be okay with the bots only checking articles under the 'category:astronomy' tree. Probably what the bots should be doing is checking for citation urls that point to the bibcode link and replacing that with the bibcode parameter. Praemonitus (talk) 01:31, 22 December 2025 (UTC)

  Yes, this is why the proposal specifically calls out cases where the ADS page adds value. In my experience ADS add value for older astronomy sources but not across the entire spectrum of wikipedia. The proposal is not about removing bibcodes. My sole goal is to stop edits that add them when they are just noise. Johnjbarton (talk) 01:57, 22 December 2025 (UTC)

• Oppose that wording. Bibcodes are generally useful, as discussed above, so should generally be included. But the specific problem here is what counts as 'unique information'? Every ADS entry provides things like citation counts, links to open access or preprint versions, links to other sources cited by or that cite this one etc. None of those are available via the DOI, so there's always unique information. This wording would appear to allow every Bibcode, which doesn't seem to be the intention and would be pointless to specify. Modest Genius ^talk 13:49, 8 January 2026 (UTC)

  As far as I can tell the above discussion does not support "bibcodes are generally useful". On the contrary, only a couple of editors can find a use for bibcodes and only in articles related to astronomy. Your claim that "every ADS entry" provides links to various bits of information does not match the examples I pulled out of an article. For example the information on 2005JRScT..42...84B is a dead link and some a list of 17 citations out of 45 given by the DOI link or 104 given on Google Scholar. The bibcode seems to only give citations within the ADS system. This matches my experience: the ADS system is fine for their core audience in astronomy but the system was not designed and does not support useful content for the breadth of Wikipedia. Johnjbarton (talk) 17:16, 8 January 2026 (UTC)

  Of course you can cherry-pick a broken entry, but even your example does provide 'unique information' that isn't available at the DOI - specifically the citations, references and metrics tabs. Whether that's enough to qualify under the proposed wording is exactly the problem I'm pointing out. One dead link does not make a database useless. I agree ADS is most useful for its core subjects of astronomy and planetary science (and is currently expanding to geophysics as part of the SciX rebrand) - perhaps restricting bibcodes to those areas would be a more sensible guideline. Modest Genius ^talk 13:47, 9 January 2026 (UTC)

  It's not "cherry picking" when the majority of the bibcodes added by the bot contain no new information. The bot currently adds bibcodes completely indiscriminately, which includes references to e.g. history, mathematics, or biology papers where the ADS database includes only the bare minimum of metadata and a tiny subset of the citation graph. If Wikipedia readers click these links, they get to a page which does not help them to locate the source or learn anything new about it. Even if Wikipedia readers don't click these links, they add a lot of clutter, because they are illegible identifiers consisting of distracting random letters.

  "perhaps restricting bibcodes to [astronomy and planetary science]" – That is more or less what is being suggested. The part people are annoyed at is the indiscriminate low-info bot-added bibcodes all over Wikipedia. I don't think anyone here has a problem with either (1) bibcodes added where ADS contains a scan of the paper and/or other related documents (e.g. code, data, appendices, extra figures), or (2) bibcodes added with human-editor support on articles about astronomy or related fields. –jacobolus (t) 19:25, 9 January 2026 (UTC)

  That is more or less what is being suggested. No, that is not the wording being suggested. I understand your concern with bibcodes, but this proposal would not fix it and could do immense damage in the fields that do use bibcodes extensively. Depending on how the phrase 'unique information' is interpreted, it could either continue the current situation, or deprecate all bibcodes, or anything in between. The proposed wording is far too vague and open to personal interpretation. Modest Genius ^talk 12:39, 12 January 2026 (UTC)

  could do immense damage – Stopping a bot from automatically adding content-free bibcodes is not going to "damage" anything. Bibcodes are not going to be "deprecated". –jacobolus (t) 16:44, 12 January 2026 (UTC)

  I tried to make a small scale proposal that would allow the important cases. How can we alter the proposal to address your concern while retaining the widely-agreed positive aspects? Johnjbarton (talk) 18:07, 12 January 2026 (UTC)

• Oppose: Bibcodes provide an alternative lookup when doi fails. In addition, I've found the former to be useful for direct author lookups in NASA ADS. An additional benefit is that the bibcode can provide photo-scanned images of (usually older) papers. Praemonitus (talk) 17:40, 22 February 2026 (UTC)

  @Praemonitus This comment is not responsive to anything that has been written in the above discussion, and is not related to the specific bibcodes being discussed here. –jacobolus (t) 18:27, 22 February 2026 (UTC)

  @jacobolus: I disagree. The wording puts the burden on the citer to demonstrate that a bibcode is needed, which is a negative connotation that will tend to have it unnecessarily removed. I'm saying that it almost always does provide unique information and/or a necessary backup for document access. There's no need to make a special rule for the exceptions that can lead to abuse. Praemonitus (talk) 18:36, 22 February 2026 (UTC)

  Did you look at Johnjbarton's list of exmaples, or at the numerous other non-astronomical bibcodes added by the bot which have next to no meaningful information when you click through? –jacobolus (t) 18:37, 22 February 2026 (UTC)

  If you dislike the wording of Johnjbarton's proposal, there could certainly be other concrete ways of cutting the number of vacuous bibcodes. One obvious one would be to just disallow bots or script-assisted humans from adding bibcodes unless they include a full text scan. That would take care of at least 99% of the problem. –jacobolus (t) 18:41, 22 February 2026 (UTC)

  I don't see this as a problem. We could debate this endlessly, but I've used bibcodes extensively, including on non-astronomy articles. I'm satisfied with my position. Praemonitus (talk) 12:47, 10 March 2026 (UTC)

  If you add bibcodes you find relevant as a single human editor, I also don't think anyone has a particular concern (if they do, it can be worked out on a local talk page). The complaint is about a bot adding tons of bibcodes with zero/marginal value to articles all over the project. –jacobolus (t) 16:28, 10 March 2026 (UTC)

Meta-comment: As this discussion was opened more than two months ago and has become fairly quiet, it might be a good idea if somebody could formally close it? Gawaon (talk) 18:18, 22 February 2026 (UTC)

• Oppose: as mentioned by others, bibcodes are a useful and standardised system for astronomical (and related) citations in the NASA ADS system. They are useful, or at least widely-used, beyond the cases where only ADS provides access to a paper, often where a doi and/or preprint also exists. If you want to stop them being sprinkled in citations in other fields where they are not a standard and not useful, then come up with a proposal for that, rather than all-but-preventing their use in astronomical articles. Lithopsian (talk) 21:38, 26 February 2026 (UTC)

  "all-but-preventing their use in astronomical articles" – I really don't understand how you inferred that, since it is almost diametrically opposite to what has been said here. –jacobolus (t) 22:59, 26 February 2026 (UTC)

  How would you feel about an alternative proposal which only said that bots or tool-assisted humans should not indiscriminately add bibcodes to articles unless they include a full text scan, but otherwise left decisions about explicit manual addition of bibcodes to local consensus? –jacobolus (t) 23:02, 26 February 2026 (UTC)

• Oppose: as mentioned before bibcodes are useful, not just in the case of older astronomy sources, but in my experience at least for any astronomy source, as easy accessible different formats are provided. The current proposal wording seems not adequate and as also mentioned the burden should not be put on the citer to demonstrate that a bibcode is needed.

Stevinger (talk) 00:26, 27 February 2026 (UTC)

The burden for additions is always on the editor who adds content. Johnjbarton (talk) 03:53, 27 February 2026 (UTC)

I agree. But we had that example above added by WhatamIdoing. If Alice adds a reference with bibcode as only source and then later Bob adds a different identifier. Then it should be the burden of Bob or another person later on to show the bibcode gives no additional information anymore and can be removed or not remove it. Stevinger (talk) 14:28, 27 February 2026 (UTC)

Note: Many of the replies here may be due to a post on Wikipedia talk:WikiProject Astronomy in which user:Headbomb characterized this discussion as "a proposal to effectively ban Citation bot from adding bibcodes", which seems to me like an intentional mischaracterization. YMMV. @Johnjbarton, maybe we can come up with a more clearly worded proposal which makes it clear that astronomy-specific articles can choose to add as many bibcodes as they feel like it, if there is a strong local consensus that even the empty ones provide value to that particular audience. –jacobolus (t) 01:04, 27 February 2026 (UTC)

I don't know how to make a clearer proposal. Any useful bibcode is allowed under my proposal, astronomy or not. I am unable to understand why adding non-useful bibcodes is allowed. If a URL, DOI, or wikilink pointed to useless trivia I believe it would be deleted. Why bibcodes are special eludes me. Johnjbarton (talk) 03:52, 27 February 2026 (UTC)

The question is nevertheless how a bot (Citation Bot or any other) would know whether a bibcode is useful? If that's hard for a piece of software to decide, then Headbomb's comment/warning seems essentially correct. Gawaon (talk) 08:44, 27 February 2026 (UTC)

I would recommend that the bot not add bibcodes unless ADS has a scan of the article, except possibly to wikipedia articles which have opted in (e.g. ones directly about astronomy). –jacobolus (t) 15:36, 27 February 2026 (UTC)

I wanted to clarify I did not read what Headbomb wrote and stumbled across this topic differently. Telling people coming in many of them did not form their opinion on their own might not easily convince them of the opposite, but YMMV. Also strong local consensus should not be a requirement to use an identifier. If you are, let's say in geophysics and add a bibcode because it provided you with a benefit in information and a third of people agree, the other two thirds (mostly not using bibcodes) should not remove it because it is a bibcode as sole reason.

The problems seem to be:

• You need to make sure people agree what a useful bibcode is. Benefit of a bibcode eluding someone should not be a reason to remove it.
• A bot needs to understand somehow whether a bibcode is useful as given above. The priority needs to be that useful bibcodes are never removed, especially the ones inserted by hand. If then a maximum of agreed upon not useful bibcodes are removed this would be perfect and help out people asking here for a change.

If these two points are covered, I assume you will get all the support.

I am writing this because I liked the previous sentence 'I tried to make a small scale proposal that would allow the important cases. How can we alter the proposal to address your concern'. But people are worried that the concept changes to 'somehow it will work out' and that will probably not be successful. Stevinger (talk) 15:35, 27 February 2026 (UTC)

Thanks! But I can't figure out how to apply your suggestions.

• "You need to make sure people agree what a useful bibcode is." As far as I can tell those who oppose do not care whether or not the bibcode is useful, the possibility that it might be useful is sufficient. My proposal already gave a criterion, that "the linked page provide unique information unavailable through other links in the citation." Is there a better criterion?
• "A bot needs to understand somehow whether a bibcode is useful as given above." Indeed! The observed fact, by both those who oppose change and those who seek it, is that the bot is inadequate to determine if the bibcode adds value. The bot is not currently good enough: why do we allow it? (My proposal says nothing about bots because in my opinion the issue is quality of content.)
• "The priority needs to be that useful bibcodes are never removed, especially the ones inserted by hand." My proposal only talks about addition.

As I reflect on the posts above it seems clear that editors who work on astronomy articles want bibcodes because they are invaluable in some cases even if they fail in others. I think this point of view places the burden on the wrong set of editors. An editor who adds content is responsible to the community for that addition. Part of that burden of responsibility is to not waste other editors time with links that have no value, even if a bot makes that easy.

I appreciate your suggestions but I don't know how to make a better proposal. I want bibcode links to work for us. Bibcodes link a lot of useful older astronomy sources: let's use bibcodes for the cases where they add value. Johnjbarton (talk) 17:52, 27 February 2026 (UTC)

Nobody is suggesting that we not use bibcodes for the cases where they are useful. But you keep insisting that the existence of some useful bibcodes is a reason to continue automatically spamming the useless ones to our references. The distinction between useful and useless seems clear to me: bibcodes with scanned article copies are useful; bibcodes with only metadata are useless (because that metadata is or should be already in our own references). And in other cases where this misses some other usefulness of a bibcode they can be added manually rather than automatically. What is the obstacle to blocking the automatic addition of no-scan bibcodes? —David Eppstein (talk) 18:38, 27 February 2026 (UTC)

I suppose the question is, how does the bot determine whether a bibcode that it happens to add contains a scanned article copy or not? I imagine that at present it only constructs the relevant URL. Cheers, SunloungerFrog (talk) 19:26, 27 February 2026 (UTC)

Such bibcode metadata pages include links like articles.adsabs.harvard.edu/full/«bibcode» and .../pdf/«bibcode» from their right-hand sidebar, under "FULL TEXT SOURCES" / "ADS". For instance, the only obituary I know of Laurence Patrick Lee (Bibcode:1985SouSt..31..221R) was published in Southern Stars, the journal of the Royal Astronomical Society of New Zealand, and is only available online through ADS, with scan at
.../full/1985SouSt..31..221R or .../pdf/1985SouSt..31..221R.
In this case, it's essential to include the bibcode in the citation, which is why I manually added it. My understanding is that citation bot gets its data from (among other places) the ADS database, which can presumably be directly queried about whether there is a scan available. –jacobolus (t) 19:47, 27 February 2026 (UTC)

If I understand correctly, the obstacle is that some astronomers would prefer to always see a bibcode wherever it exists, because they are accustomed to using ADS for most of their work. I think a special article-specific opt-in can handle this kind of case. (Or just a non-bot tool that a human who loves bibcodes can run where there is local consensus supporting it.) –jacobolus (t) 19:28, 27 February 2026 (UTC)

Thanks! I appreciate you discuss the points, but I have to admit your answer is disappointing.

• 'As far as I can tell those who oppose do not care whether or not the bibcode is useful' Well, this is the second time I am answering and I am one of the persons who oppose. If you assume others don't care instead of looking for common improvements things are of course difficult. 'Is there a better criterion?' I said you need to make sure people agree what a useful bibcode is. This is easier if people that give ideas are listened to. In the discussion above with Modest Genius Jacobolus gives a better description of what you likely meant with "the linked page provide unique information unavailable through other links in the citation." and Modest Genius disagrees that this is 'not the wording being suggested' in the proposal. So the proposal is likely too vague and points to improve it are given. (If everyone assumes they know what useful means in this context and this is the only possible way then this won't work, and I am not speaking of you Johnjbarton in this case.)
• 'My proposal says nothing about bots because in my opinion the issue is quality of content.' I do understand what you actually mean, what is correct, but this statement is again wrong, regarding you directly write 'The proposal would apply to bots as well.' in your proposal.
• 'My proposal only talks about addition.' This is correct, but if you follow up by 'In my experience DOI has effectively replaced bibcode. The bibcode typically only shows the same info as the citation.' this can mean a lot of things. In my experience bibcode always provides more information than DOI when both have relevant data and are not just dead links (or on either side). This means one could remove DOI in cases both give relevant data, but this seems not necessary.

You write what you want is that 'bibcode links to work for us', but you keep repeating things like 'link a lot of useful older astronomy sources'. Several people said this is not giving a correct picture and thus not helpful, and let's people doubt you ever used bibcodes. Again, I appreciate a discussion, but if not even a first proposal can be improved by arguments already raised during the discussion and by a description how you would want to change the bots I am doubting this will work. At the moment it feels like 2 steps forward and then randomly 1-3 steps backward. I am sorry, but the current proposal text can mean a lot of things as emphasized by half of the discussion participants and is thus not acceptable. Not knowing 'how to make a better proposal' is unusual when some lines of input is already given.

Addendum: I just realized Jacobolus gave as last entry (by time it was added) a description of how bots could be changed (above). Another thing that would improve a proposal, especially if it is not only presumably possible to be incorporated. Stevinger (talk) 23:42, 27 February 2026 (UTC)

There is a discussion at Help talk:Footnotes § Tag or template preference which may be of interest. -- Beland (talk) 05:28, 18 December 2025 (UTC)

Per this, would there be any objections to changing the use of {{reflist}} to <references /> on this page? The native mediawiki markup makes sense as the default here, though this example doesn't use LDR which would make the case stronger. Mike Christie (talk - contribs - library) 18:39, 31 December 2025 (UTC)

I don't see any reason to make the change, does it cause a problem as it does with LDRs? -- LCU ActivelyDisinterested «@» °∆t° 18:45, 31 December 2025 (UTC)

On a page that is hitting the PEIS limit, reflist makes the situation worse, unlike the tag, because it's a template. I don't know if there are other reasons beyond those two. My suggestion here was prompted partly by discussions at Template talk:Reflist, where for example Anomie commented that changing between the two should ideally be a cosmetic edit. If that's the case (and given there are a couple of reasons to prefer the tag) it seems to me this page should give the tag as the example. Mike Christie (talk - contribs - library) 19:20, 31 December 2025 (UTC)

This template adds little to the PEIS limit, it just happens to be at the end of articles and so is after the PEIS limit being hit. The problem in every case is the content of the article, not this template. -- LCU ActivelyDisinterested «@» °∆t° 19:58, 31 December 2025 (UTC)

I changed the example on this page, as the other discussion seemed to generally prefer the tag. -- Beland (talk) 20:29, 1 January 2026 (UTC)

@Beland WP:REFLIST shows <reflist /> twice, that's probably not what you intended with your change? Johannes Richter (WMDE) (talk) 09:28, 2 January 2026 (UTC)

Huh, I thought I saw that stay the same when I previewed. I must have not refreshed before publishing. Fixed now; thanks for catching that! -- Beland (talk) 17:41, 2 January 2026 (UTC)

There was an RfC in Spring 2025 that concluded that titles of all sources within a given article should all use the same capitalization rule (e.g. title case, or sentence case); and that the capitalization that individual sources use for themselves should be ignored. That decision is documented in this guideline as:

• WP:CITESTYLE: "Nearly any consistent style can be used [for citations]. Preserving the capitalization style that each individual source uses for itself is not considered a consistent style.".

But this guideline also has:

• WP:CITEVAR: "... you may choose whichever style you think best for the article, except inline parenthetical referencing. Follow the citation style's preferences for whether to use title case, sentence case, etc.,..." (emphasis added)

The wording is not very clear for a couple of reasons:

1. Searches for "title case" will lead editors to WP:CITEVAR which says ... choose whichever style you think best.... That search will not lead editors to to more relevant "Preserving the capitalization style ..."
2. The wording "Preserving the capitalization style that each individual source uses for itself is not considered a consistent style." is a overly poetic, and does not plainly state the guideline to readers. Better would be something like All source titles within an article should conform to the same capitalization rule, such as title case or sentence case (however, using the capitalization that individual sources use for themselves is not acceptable).

I'm not bringing this issue up randomly: I've had to cite this guideline three times in the past few months in FA reviews, and it is very difficult to locate the RfC's guidance anywhere in the MOS or guidelines. It takes me a long time to find it. The wording of the guideline should include search-friendly phrases that help editors find the guidance. Thoughts? Noleander (talk) 22:48, 31 December 2025 (UTC)

I do not believe that the conclusion you state ("that titles of all sources within a given article should all use the same capitalization rule (e.g. title case, or sentence case") was the declared consensus of that RFC. It merely declared that following the publications' own capitalizations for individual references is not a consistent citation style. In particular, the closing statement included "This consensus is not a consensus that a uniform capitalization, such as sentence case or title case, should be imposed."

I tend to personally follow a citation style in which books and periodicals use title case and that individual articles, chapters, or other contributions within a book or periodical use sentence case. I believe that rule to be consistent and I do not believe that the RFC outcome says anything to the contrary. —David Eppstein (talk) 23:37, 31 December 2025 (UTC)

Thanks for the clarification of the outcome of the RFC. The fact remains that the guidance Contained in this page is extremely confusing... I cannot tell from the this Wikipedia page what the actual guidance is. If what you say is correct, then the page shoukd be updated to clearly state that. If this is a fairly minor aspect of the guidance, perhaps it can be stuck into the existing footnote on this issue. Otherwise future editors will have no idea what the guidance is. Noleander (talk) 00:27, 1 January 2026 (UTC)

I believe the wording "All source titles within an article should conform to the same capitalization rule" is too strong to be realistic. Yes, in an ideal article that will be the case, but most articles won't comply with that ideal since references are added by editors on a case by case basis. When improving articles, I often try to clean this up a bit – say towards a policy of "title case for books, sentence case for shorter articles", which corresponds to what David Eppstein said and often matches de facto practices quite well. Perfect correspondence to title or sentence case for all references is often not realistic without lots of editorial time spent on it. I think the result of the RfC was simply: Editors may clean this up, by converting everything to either title or sentence case or some other consistent rule; other editors may not revert this by saying "but the source itself uses a different case". That was all the RfC was about and that was its outcome. But editors don't have to clean articles up, and in that case the article will simply remain in an imperfect state regarding the casing style. Which I guess is the case with 99% of our articles with more than ten references, if one only considers "title case everywhere" or "sentence case everywhere" as consistent. Perfection is hard to reach in an imperfect world. Gawaon (talk) 03:39, 1 January 2026 (UTC)

Thanks for the further clarification of what the RfC was saying. But I think you're just proving my point: the guideline is not clearly explained in this project page. You wrote "Editors may clean this up, by converting everything to either title or sentence case or some other consistent rule; other editors may not revert this by saying 'but the source itself uses a different case'. ".... Which is fine. But that's not what this project page says. And it's not what editor Eppstein said above. Apparently, people that participated in the RFC cannot even agree on what the outcome of the RFC was. I repeat my assertion that the guideline on how to capitalize source titles is unclear and this page should be improved. I have no opinion on what the guideline should be, But whatever it is, it should be clearly stated in this project page. Noleander (talk) 05:40, 1 January 2026 (UTC)

You appear to think that the outcome of an RFC must always be a rule telling you exactly what to do in all circumstances. That is not the case, and not the case for this RFC. We have no rule other than to use consistent casing within a single article, and no definition of what consistency means beyond to use common sense and to not try to follow publisher casing. —David Eppstein (talk) 06:54, 1 January 2026 (UTC)

@Noleander, does this still need work? WhatamIdoing (talk) 01:19, 27 January 2026 (UTC)

@WhatamIdoing - Yes. But I'm kinda burned-out on MOS. Every time I come to the MOS Talk page to point out ambiguities or opportunities for clarification, I get fire-bombed by very experienced editors (who should know better) telling me I'm a moron. It is a bit discouraging. So, yes the guidance for source title capitalization guidance is still unclear & contradictory. Whatever happened to the notion of rolling-up our sleeves and working together? If someone wants to improve the clarity, I'm happy to assist. Noleander (talk) 01:27, 27 January 2026 (UTC)

@Noleander, does this edit address all of your concerns? WhatamIdoing (talk) 18:37, 3 February 2026 (UTC)

Yes, that looks much better. Thanks for doing that. Noleander (talk) 20:06, 3 February 2026 (UTC)

You're welcome. We'll see if anyone volunteers to have another discussion about this or if the edit sticks. WhatamIdoing (talk) 03:21, 4 February 2026 (UTC)

See the initial discussion at the Help Desk

I'm seeking clarification on whether to remove dead links that have been tagged for longer than 2 years and tag with {{cn}} if applicable per WP:DEADREF or to keep them per WP:KDL.

As of posting, three editors in the help desk have argued to keep the dead links. This discussion was the only recent discussion to support DEADREF that I found.

What is the best option for this? Thanks. Ecourter (talk) 19:15, 6 January 2026 (UTC)

Honestly, what's the hurry or point? Tag as dead and leave it for reference or archival pursuit. Add an archive.org or archive.today archival if one exists. Don't just yank a latent reference point someone can use later for arbitrary tidyness. If no attempt is made to archive/correct, don't gnome for the sake of gnoming! — Very Polite Person (talk/contribs) 19:22, 6 January 2026 (UTC)

Then what is the point of having step 6 on DEADREF? If the consensus is to leave dead links alone, shouldn't that be removed or at the very least rewritten? Ecourter (talk) 19:34, 6 January 2026 (UTC)

Sure, if literally all options are exhausted. I'm just saying there's never any imperative or time sensitive or semi-automated "let's knock out 1K" of these for urgency. — Very Polite Person (talk/contribs) 19:38, 6 January 2026 (UTC)

I should have phrased myself a little better. I would not immediately remove the dead link, but I would do the checks and see if I can find an archive or alternative, and then remove the dead link and CN tag it per step 6. I have yet to seek out these dead links, I've only stumbled across them. Trying to figure out how to do all that semi-automatically would be out of my depth. Ecourter (talk) 19:49, 6 January 2026 (UTC)

I may be a bit biased as I've found other supplemental sources by scratching at what I assumed were dead links; suddenly putting another keywork into google with "some phrase text" and things like filetype:pdf ends up finding other content, and then I find a backway/alternative off of references or footnotes or some other errata that invoked (but did not share more data on) the original dead link. I was kicking myself like that ready to give up on a particular document more than once that I knew existed but couldn't find. Then I start searching sideways and find a relevant piece of data from the missing source, but mentioned elsehwere in a roundabout way.

If you find one, they're probably rare these days as so many people are obsessive archivers. Trust me; the internet's old enough now with us as a nexus of data cross-referencing (especially as so many use us to bootstrap/start their research) that there's probably a goofy way to get at dead content. Once I found one in Archive.org as a book I could borrow, that didn't even exist where I could buy it online anywhere (and I tried). — Very Polite Person (talk/contribs) 19:54, 6 January 2026 (UTC)

In my eyes, there is only three reasons to remove a dead link from an article:

1. The information being cited is removed

2. The source has been replaced by a better one

3. We can tell that the information would never have been at that location.

Removing links and not replacing them is just a loss of information. Lee Vilenski ^{(talk • contribs)} 20:25, 6 January 2026 (UTC)

It depends on the information. If you think the information is accurate and that another source is likely to exist (even if you haven’t found it) - leave it. If you think it inaccurate or unlikely that another source exists - remove it. Be stricter in BLPs than in other articles. Blueboar (talk) 20:48, 6 January 2026 (UTC)

The best thing to do when possible is to find a replacement for the dead link, either an archived copy or an alternative link for the same document. For print references with unrecoverable dead links, the link should be removed from the reference but the reference should not be removed; there is no problem with references to offline print publications. It is only when the reference existed only as a web page and that web page cannot be recovered that it needs to be replaced by some other reference. Even in those cases, I would prefer leaving the deadlink in place than replacing it by {{cn}}, until a replacement can be found, because often the deadlink can provide useful information that might help some other editor find an appropriate replacement. —David Eppstein (talk) 21:05, 6 January 2026 (UTC)

@Ecourter, have you been adding the |fix-attempted=yes parameter to the {{dead link}} template? This lets editors know that you've done all the hard work. WhatamIdoing (talk) 01:22, 27 January 2026 (UTC)

Hi, Is there a way to cite multiple references within a single cite web / cite book etc section of code? I have a few instances where Reference A (an out-of-print book) is considered a reliable source, and the same data has been transcribed to Reference B (an enthusiast site for the same topic, so wouldn't meet WP:RS on its own merits). Ordinarily for Wikipedia purposes it would be correct to only include Reference A, but I also want to link to Reference B as a more convenient way to get the same information. So, what sort of code should I use to say "this should be source A, but the same info is also available at less reputable source B which references A"? Anothersignalman (talk) 09:39, 8 January 2026 (UTC)

@Anothersignalman, I think what you are talking about is a convenience link; see also the essay Wikipedia:Convenience link. If you search for "convenience" in {{Cite book}} then it seems you can include such a link in the url or title-link field. However, you have to be sure that your Reference B is not infringing the original rights-holders' copyright of Reference A. If it does, you mustn't link to it. I hope that's somewhat helpful. Cheers, SunloungerFrog (talk) 09:52, 8 January 2026 (UTC)

Thanks for that. Yes, that seems like the right sort of thing. Looks like Wiki doesn't distinguish between a proper URL and a convenience URL; I was hoping for something like a Convenience-URL = field in the same way that URL and Archive-URL are separated. Possibly some code challenges since (for example) page 100 of the book might link to site/page1.html, page 102 might link to site/page337.html etc, and just linking to the overall site might not be sufficiently convenient. The book side is solved with rp|37. but that doesn't really work for websites as far as I know? Re copyright, in one instance I had permission from the author of the book (before they died) to transcribe the content, in another it's between two expert self-published sources so I presume they'd worked out something between them. Is that a fair assumption? Anothersignalman (talk) 10:23, 8 January 2026 (UTC)

If you were to use a {{Shortened footnote}}, it seems that for each instance of an inline citation, you can add a URL for a loc= or p= parameter. See Template:Sfn § Adding a URL for the page or location On the copyright front, I wouldn't make any assumptions, and if there's any doubt, I would ask someone more qualified than me! Maybe ask one of these admins for specific advice on the sources / copies in question. Cheers, SunloungerFrog (talk) 10:47, 8 January 2026 (UTC)

I recommend User:SMcCandlish/How to use the sfnp family of templates as it does a great job of explaining how to handle awkward csses.

(btw, {{sfnp}}, which puts the date in parentheses like 'normal' CS1/2, is "prettier" than {{sfn}} for that reason. Otherwise they are identical.) 𝕁𝕄𝔽 (talk) 16:29, 8 January 2026 (UTC)

Ooh that is useful - thanks for linking the essay! Cheers, SunloungerFrog (talk) 16:43, 8 January 2026 (UTC)

You can always include information after the cite but before the closing reference tag.
<ref>{{cite book |last=Last |first=First |title=title |date=year}} [https://website.com/translation A translation is also available]</ref>. -- LCU ActivelyDisinterested «@» °∆t° 16:02, 8 January 2026 (UTC)

Oh, I like that idea! Thanks :) Anothersignalman (talk) 01:20, 9 January 2026 (UTC)

Alternatively, you might be able to use via= ? But you would have to be sure that the transcription is faithful, meaning that you have to have read the original text yourself. Is the original book on archive.org? (or, second best, Google Books? 𝕁𝕄𝔽 (talk) 16:22, 8 January 2026 (UTC)

Depends what you mean by faithful. In the first case, it's not letter-for-letter because I expanded acronyms and shorthand to full words as outlined at the start of the book (e.g. "ILO" to "in lieu of", and plenty of technical terminology), and there was an errata the author kept up-to-date until his death that I handwrote into my copy. (I also had to tape the spine back together after I'd finished copying it.) I don't know who is handling the estate, I assume it'll be a few years before it's resolved. That book is not available online except as the transcription I made, as far as I know. For the second example, between the two Expert:SPS's, I have no idea, but even if it is available online I think I'd like a copy for myself. Anothersignalman (talk) 01:27, 9 January 2026 (UTC)