I, for one, miss Citation Bot

I don't understand what she's done wrong exactly but I miss her and I wish she wasn't in timeout. jengod (talk) 06:58, 1 March 2026 (UTC)

It wasn't intended to be permanent, merely until the significant problems uncovered in recent large unsupervised batch runs were fixed to the point of it being safe enough to run batch runs again. I haven't heard much here about progress on fixing those problems, though. —David Eppstein (talk) 08:03, 1 March 2026 (UTC)

I understand that there are problems with unsupervised batch runs but don't understand why it has been blocked from running on single pages. Last time there were problems with the bot it was only blocked on batch runs. Even with the current problems this is still a useful tool for single pages where any errors can be manually resolved.

On a broader topic, I'm not sure unsupervised batch runs are compatible with Wikimedia Foundation Terms of Use You take responsibility for your edits. --John B123 (talk) 09:05, 1 March 2026 (UTC)

"why it has been blocked from running on single pages"

Abuse by Adbuctive. Headbomb {t · c · p · b} 16:23, 1 March 2026 (UTC)

Oh I have experience with this! When he or she runs CB on my pet over-long artwork article it adds scraps from HathiTrust that I do not love and have to go in and scrape out, most annoyingly |pages=[4], 400 p but also it's not the end of the world.

I'm afraid to ask, but how many articles are touched in a case of "bulk task abuse"? Could we just pick an arbitrary number (85! 243! 12! 2,000!) and create a maximum number of Citation Bot requests per user per day? jengod (talk) 16:40, 1 March 2026 (UTC)

I would love this too. Start at 243 and titrate from there? Why aren't we discouraging Adbuctive's use rather than all use? – SJ + 03:06, 2 March 2026 (UTC)

+1. I have not reviewed the above-mentioned batch runs but I know that error rates tend to be higher with batch runs over broad categories etc. Batch runs are more effective when targeting specific issues. Perhaps we need a tracking category to target whatever it is that those batch runs were trying to achieve? But that can be discussed after restoring the bot in a more limited fashion. Nemo 17:08, 2 March 2026 (UTC)

Whitelisting categories (many are already whitelisted) can be a solution. Another would be to restrict who can trigger the bot on worklists. Yet another would be to restrict specific users to 10 bot edit per hours. Headbomb {t · c · p · b} 17:12, 2 March 2026 (UTC)

Many people miss the bot! I, along with some other users, left Fs in the Discord server 🙁 - OpalYosutebito 『talk』 『articles I want to eat』 19:48, 4 March 2026 (UTC)

Yes. I agree with above mentioned solution. Time to open to bot again fully. And if mistakes are made again, we can simply block it again until those issues are fixed.BabbaQ (talk) 07:43, 12 March 2026 (UTC)

Wait, CB is blocked? Why? Isn't CB the most iconic WP bot next to Lowercase sigma III, Internet Archive Bot, and Twinkle? VidanaliK (talk to me) (contributions) 16:30, 12 March 2026 (UTC)

@VidanaliK apparently Citation Bot had been making mistakes, which led to it getting blocked, but I agree that it is very iconic 🙂 - OpalYosutebito 『talk』 『articles I want to eat』 17:23, 12 March 2026 (UTC)

Some edge cases of Citation template situations were not being handled according to expectations (really hard to say from my read of the discussion that these were actually significant errors) and also made worse because one particular editor was insisting on running massive batch jobs without any review or going back to fix. Vast majority of people using the bot were running on single pages at a time and checking/fixing the results. But now no one can use it. And the gadget only works on smaller articles or one section at a time.   — Chris Capoccia 💬 15:08, 14 March 2026 (UTC)

I miss the bot. Artoria2e5 🌉 07:22, 17 March 2026 (UTC)

@David Eppstein @Headbomb I did some digging to try to find the conversation that led to CB getting blocked but could not. Could y'all point me towards it? I'm surprised to hear multiple people argue that this is Abductive's fault, and yet have the bot blocked. CaptainEek ^{Edits Ho Cap'n!}⚓ 02:37, 20 March 2026 (UTC)

See User talk:Citation bot/Archive 44#Blocked; huge number of bad work= additions above. Headbomb {t · c · p · b} 02:49, 20 March 2026 (UTC)

Since the bot was blocked plenty of improvements have been made to multiple previous problems with the bot. I think it is time to unblock.BabbaQ (talk) 16:52, 20 March 2026 (UTC)

I only use the bot on single pages - how can I formally ask for it to be unblocked at least for single pages? Chidgk1 (talk) 09:04, 23 March 2026 (UTC)

Citation expander: automatically expand and format citations using Citation bot. Moxy🍁 16:46, 23 March 2026 (UTC)

.@Moxy: I still get a bot blocked or not enabled message if I use Citation expander. --John B123 (talk) 20:21, 23 March 2026 (UTC)

Are you trying to use the bot or using the edit window?

Moxy🍁 20:24, 23 March 2026 (UTC)

Apologies, I was using the 'Expand citations' link in the l/h Tools menu. It works OK from the button on the edit window. --John B123 (talk) 20:43, 23 March 2026 (UTC)

Replace hardcoded nbsp with space

The bot stays away from the |quote= since it is often formatted by people for some specific reason. AManWithNoPlan (talk) 19:15, 12 July 2025 (UTC)

@AManWithNoPlan: Then they should use an explicit   to indicate intention if that's ever the case. Same for the other hardcoded special whitespace like thinspaces. Everything not explicitely coded should be convert to a plain space. Headbomb {t · c · p · b} 18:05, 14 July 2025 (UTC)

@Redalert2fan: why the no fix? WP:NBSP is clear about hardcoded instances being unacceptable. Headbomb {t · c · p · b} 18:16, 30 March 2026 (UTC)

Since the bot doesn't edit quotes on purpose, this is not something we automatically fix. Everywhere else it should be replaced by the bot. Redalert2fan (talk) 18:21, 30 March 2026 (UTC)

So to clarify, understand what you are saying about NBSP. but a won't fix since the bot doesn't edit |quote= by design. Redalert2fan (talk) 18:54, 30 March 2026 (UTC)

Citation bot removes "|chapter=" when "|trans-chapter=" exists, resulting in CS1 error

Alexanderino, you had the |chapter= parameter duplicating the |title= parameter in this citation (which I fixed here). Agree that the check for requisite additional parameters should happen before removing duplicate parameters so as to avoid this sort of error, but it's always a slog to capture all the potential GIGO cases.

Incidentally, the three cites in that section to L'auto should have been calling {{cite periodical}} or {{cite news}} instead of {{cite book}}. I fixed that here, additionally specifying issue dates and page numbers so the references can still be verified if the Gallica url stops working. Can you check the other citations to ensure the full information is included? Folly Mox (talk) 12:25, 7 February 2025 (UTC)

Thank you very much for your clarification and help. I will take a look once I've had some sleep. Alexanderino (talk) 12:31, 7 February 2025 (UTC)

Patch: PR 5506 Redalert2fan (talk) 19:55, 30 March 2026 (UTC)

Artistdirect duplication

Should have been (indirectly) solved by PR 5497 --Redalert2fan (talk) 18:23, 30 March 2026 (UTC)

Purge pubmed URLs

Per the RFC. See also User_talk:Citation_bot/Archive_41#Yeet_PMID/OCLC_links_from_citations. Headbomb {t · c · p · b} 05:59, 27 August 2025 (UTC)

Patch: PR 5512 --Redalert2fan (talk) 18:50, 31 March 2026 (UTC)

Cosmetic bot converting title= to chapter= in cite encyclopedia

Solved with PR 5503 --Redalert2fan (talk) 14:06, 30 March 2026 (UTC)

changes |title= to |chapter= in {{cite encyclopedia}} template

Same as § Cosmetic bot converting title= to chapter= in cite encyclopedia above —David Eppstein (talk) 19:05, 22 January 2026 (UTC)

Fixed with PR 5503 --Redalert2fan (talk) 14:10, 30 March 2026 (UTC)

Adds duplicative text not shown by the source

• The very last edit the bot made before it was blocked introduced this error, which I think is of a type not reported before. Abductive (reasoning) 00:39, 2 February 2026 (UTC)

Anyway, something is messed up. Abductive (reasoning) 03:02, 2 February 2026 (UTC)

I Can't tell you exactly which patch fixed this, but one of the recent ones did. Can't replicate the bug anymore. --Redalert2fan (talk) 13:33, 2 April 2026 (UTC)

|title= parameter removed (April 2026)

Moxy suggests this edit. |title= parameter should be re-added to fix Category:CS1 errors: missing title. Achmad Rachmani (talk) 09:44, 2 April 2026 (UTC)

Your fix |title=iran-regime-intelligence-irgc-war is not a useful fix. Do not do that. That 'title' is wholly meaningless. If the bot cannot determine a correct title (in this case: "U.S. intelligence says Iran's regime is consolidating power"), deleting the parameter / value pair |title=... (where the ellipses are present in the wikitext) or just deleting the ellipses is a good fix because it should draw the attention of a human editor who can add the correct title. Inserting ... or iran-regime-intelligence-irgc-war masks the error message so no one knows that the citation has a bogus title.

—Trappist the monk (talk) 13:42, 2 April 2026 (UTC)

To be honest, I don't think the bot tried to determine a new title at all for this one. I can't say for sure but it either removed the bad title because it is specifically recognized as a bad title, and after this is done there is no trigger again to try to add a correct title leaving an empty parameter. That or it just strips ellipses from titles in general. I would need to dive a bit deeper to determine what specifically activates here.

Both cases then the title= parameter becomes empty which would afterwards be deleted by the remove empty parameters logic. That logic is one of the last things to run, definitely after the part that adds new titles.

Normally I would agree that the bot probably shouldn't cause articles being added to maintenance categories, but I feel in cases like this it is justified.

Lastly I could investigate if the bot could add the correct title as well in the same edit, but just trying the citation on its own with an empty title= the bot doesn't seem to be able to come up with a title either. Improvements could be made but low priority since more pressing issues exist at the moment. Redalert2fan (talk) 22:53, 2 April 2026 (UTC)

But a very quick search: washingtonpost\.com is in the ZOTERO_AVOID_REGEX with a comment "Fails 100%". Meaning for Washington Post we never add titles via Zotero, so nothing further can be done to add a good title if we remove a bad title from a Washington Post article specifically. Redalert2fan (talk) 23:04, 2 April 2026 (UTC)

Bad title: The Wikipedia Library

Patch: PR 5521 --Redalert2fan (talk) 09:37, 4 April 2026 (UTC)

URL gets properly converted now. For this specific URL bot can't generate a title, but that is not a bug. --Redalert2fan (talk) 13:24, 4 April 2026 (UTC)

New bug

|format=pdf is added by the template so is not needed. The authors are correct. The work/publisher bug is reported above. Headbomb {t · c · p · b} 04:04, 7 February 2026 (UTC)

Umm, |last1=John |first1=Gilliland is definitely wrong at ref 40 (permalink).

—Trappist the monk (talk) 19:10, 7 February 2026 (UTC)

It mixed last and first, yes. Headbomb {t · c · p · b} 01:06, 8 February 2026 (UTC)

Investigationn: UNT Digital Library metadata provides names in "Family, Given" order but Zotero misparses them, placing the family name in firstName and the given name in lastName Redalert2fan (talk) 13:56, 2 April 2026 (UTC)

Unfortunately It's not possible to fix this on our side, Zotero would have to do that. for now will tell the bot to ignore UNT Digital Library to not introduce errors. --Redalert2fan (talk) 19:27, 4 April 2026 (UTC)

Weird removal of issues for no apparent reason?

According to Template:cite book "Issue number. This parameter is not supported by and should generally not be used with cite book. Consider that a different cite template may be more appropriate, such as cite magazine or cite journal. See Help:Citation_Style_1#Pages." Now if the template should have been converted to another type like as in the example a cite journal or magazine which support Issue= the bot could not do it, because apart from the title and date the citation lacks any data useful for that. Conversion would run normally before stripping. --Redalert2fan (talk) 15:54, 4 April 2026 (UTC)

Ah I missed it was a cite book. I don't know what those are supposed to be though. Just ignored this then. Headbomb {t · c · p · b} 15:57, 4 April 2026 (UTC)

No problem. I think it might be document "No 2 1971 Sao Paulo" displayed on this page. If you click it it will bring you to a copy of it on Yandex (Russian). I would say it would be best described as a magazine. Redalert2fan (talk) 16:12, 4 April 2026 (UTC)

Which would indeed make it to be Issue 2 of Grandes Clubes Brasileiros Redalert2fan (talk) 16:13, 4 April 2026 (UTC)

I've corrected the citation in the article. --Redalert2fan (talk) 16:30, 4 April 2026 (UTC)

doi-broken-date pn page where doi is not broken

What's the problematic bot edit exactly? Something from 2025? Highly likely the DOI was broken at that time and since got repaired. Headbomb {t · c · p · b} 00:29, 15 April 2026 (UTC)

Can confirm the bot does think the DOI is working, see . Headbomb {t · c · p · b} 00:31, 15 April 2026 (UTC)

Okay. Will raise the issue again if I see it repeated. Hawkeye7 (discuss) 01:14, 15 April 2026 (UTC)

Add 10.1257/jep to free doi

For Journal of Economic Perspectives. Headbomb {t · c · p · b} 21:57, 12 April 2026 (UTC)

PR5526. Redalert2fan (talk) 11:54, 16 April 2026 (UTC)

{{fixed}}. Redalert2fan (talk) 18:28, 19 April 2026 (UTC)

Bad title: "Making sure you're not a bot!"

Should be added to the list of bad titles, or whatever the bot does with those. Headbomb {t · c · p · b} 12:42, 12 April 2026 (UTC)

Patch: PR 5530. Has been put on the always bad data list so will also get automatically removed. --Redalert2fan (talk) 13:24, 21 April 2026 (UTC)

Welcome back!

I know you're a bot, not a person but I thought I'd acknowledge your return to activity which I just noticed occurred two dqys ago. Thanks to all of the editors who helped patch up any problems that existed. Liz ^{Read! Talk!} 02:05, 2 April 2026 (UTC)

+1 Warm Regards, Miminity (Talk?) (me contribs) 13:31, 3 April 2026 (UTC)

I would like to give the same welcome back. --BabbaQ (talk) 23:02, 6 April 2026 (UTC)

Just checked today that it is back to work. Thanks for all who helped to resolve (temporary?) the problems. I've tried maybe two weeks ago or before that but blocked at that time. Good to have the tool bvack after circa 2 months of block. Anas1712 (talk) 20:17, 7 April 2026 (UTC)

Very happy and relieved to see the bot back in the saddle; welcome back! Vestrian24Bio 10:37, 8 April 2026 (UTC)

A goat for you!

Still one of the best bots on the wiki. Man is it good to see you unblocked <3<3<3 🙂

Plasticwonder (Cat got your tongue?) 01:56, 5 April 2026 (UTC)

Blocked; huge number of bad work= additions

I woke up this morning to find another 18 bad changes to work= (duplicating publisher or in some cases replacing publisher), on my watchlist alone: . I have blocked citation bot until this is changed to stop. Occasionally when publisher= already exists the addition of work= is a good change; far more often it is incorrect. At least one of these bad edits also included the problem of adding an author who is not the actual author of a piece of text but merely the staff member who uploaded it to a web page, already reported last March in § Far too aggressive in using hidden web page metadata and apparently never fixed.

There were many more useless bot edits adding bibcodes that do not provide any useful information, and at least one undo on my watchlist of such an edit by another editor; that is not a blocking offense in my mind but see Wikipedia talk:Citing sources § Proposal to stop adding Bibcode values to citations unless they provide information for an ongoing discussion that seems to be trending against doing that.

These were all at the initiative of User:Abductive and there have been suggestions above that Abductive be topic-banned from using Citation bot. Perhaps someone should start a WP:ANI thread to discuss that. But this problematic work= issue goes beyond Abductive's edits; if it weren't them it would be someone else. —David Eppstein (talk) 20:03, 1 February 2026 (UTC)

I looked at each one of the diffs that you provided. {{citation}} assumes (always has) that in the absence of a work parameter (|journal=, |magazine=, |newspaper=, |periodical=, |website=, |work=), the source it is citing is a book so renders the citation accordingly: the value in |title= is italicized. For example, this from Natalie Crawford (one of only three working urls among the listed diffs):

{{citation|url=https://www.usafa.af.mil/News/Article/619534/2012-thomas-d-white-national-defense-award-goes-to-rand-corporation-official/|title=2012 Thomas D. White National Defense Award goes to RAND Corporation official|date=November 15, 2013|first=Veronica|last=Ward|publisher=United States Airforce Academy|access-date=2023-03-19}}

Ward, Veronica (November 15, 2013), 2012 Thomas D. White National Defense Award goes to RAND Corporation official, United States Airforce Academy, retrieved 2023-03-19

If we follow that citation's title link, we see that the cited source is not a book but rather a news article on a website. A quoted upright title is the correct format for such sources. Replacing |publisher= with either of |website= or |work= instructs {{citation}} to render a properly formatted citation:

{{citation|url=https://www.usafa.af.mil/News/Article/619534/2012-thomas-d-white-national-defense-award-goes-to-rand-corporation-official/|title=2012 Thomas D. White National Defense Award goes to RAND Corporation official|date=November 15, 2013|first=Veronica|last=Ward|website=United States Airforce Academy|access-date=2023-03-19}}

Ward, Veronica (November 15, 2013), "2012 Thomas D. White National Defense Award goes to RAND Corporation official", United States Airforce Academy, retrieved 2023-03-19

I random-sampled three of the listed articles with dead urls:

Xiaoyu Luo: dead link archive snapshot

Sharon Goldwater: dead link archive snapshot

Sylvia de Neymet: dead link archive snapshot

None of those sources are books so using {{citation}} in book mode is incorrect. For these three citations (and the Airforce Academy source), the bot, while perhaps imperfect in its execution, was correct in applying a work parameter for these sources. Perhaps that is also true for the others in your list?

—Trappist the monk (talk) 23:18, 1 February 2026 (UTC)

No, adding work=The Royal Society of Edinburgh to a citation that already includes publisher=Royal Society of Edinburgh (on Xiaoyu Luo) is not correct. It is redundant, uninformative, and misplaced because the royal society is an organization not the name of a work that they publish. If you think that this citation needs a work to satisfy some obscure formatting requirement, you are wrong and your opinion here is discountable. A dogmatic belief that a one-size-fits-all template is somehow calling this a book and formatting it wrong because of that is your belief only. And using a badly-programmed bot to enforce your wrong opinion to the exclusion of all others, grinding everyone else down with thousands of edits clogging up their watchlists, is the oppopsite of the consensus process that Wikipedia follows. This citation is merely a citation to a thing with a single-level title rather than a two-level title/work contribution/title (or whatever) pair of titles. If the citation template formats all such things as book, in a way that creates incorrect appearance of citations, then that is also a bug in the citation template but it in no way obviates the problematic behavior of Citation bot. And in this case the bot somehow conjured up the name of the work from a deadlink, so I am skeptical about the source of its information as well as about the quality of the results. An archived copy reveals that it would be reasonable (although, I believe, not necessary), to put work=Fellows. Needless to say that is not what the bot did. —David Eppstein (talk) 23:54, 1 February 2026 (UTC)

Again, Trappist is wrong on this, the US Airforce Academy is the publisher, not the website/work. Headbomb {t · c · p · b} 00:23, 2 February 2026 (UTC)

I fully support blocking the bot until these issues are addressed. The idea is to eventually allow free rein to all users to run massive jobs, which of course would exponentially increase the errors. Users would be getting blocked from using the bot every day. Abductive (reasoning) 00:29, 2 February 2026 (UTC)

I would also support a temporary indefinite block on the bot until we can address the above issues. It has caused some duplication errors in the past, especially with the work and publisher parameters. sjones23 (talk - contributions) 02:08, 2 February 2026 (UTC)

To be clear my block was intended only as temporary. It was indefinite only because I have no idea how long it might take to resolve these issues. —David Eppstein (talk) 08:38, 2 February 2026 (UTC)

That is clear, last time it was a 72 hour block that had to be changed in to an indefinite one untill fixed as well. Redalert2fan (talk) 14:49, 2 February 2026 (UTC)

Block reported on Github as well. Specific bug was already reported but nobody picked up fixing it (last check 1 Feb). Tracking added to this section. Redalert2fan (talk) 14:56, 2 February 2026 (UTC)

Is the idea really that we want to enable more large batches of unsupervised work? I'd rather the bot be available for small supervised edits than always getting banned because of the unsupervised edits run amok.   — Chris Capoccia 💬 00:06, 10 February 2026 (UTC)

Batch work is a godsend when you have targetted issues and worklists. The issue here is that Abductive constantly abuses that feature and sends the bot on unsupervised, unreviewed, untargetted random trawls. Headbomb {t · c · p · b} 00:40, 10 February 2026 (UTC)

It's also breaking {{citation|journal=}} by changing |title= to |chapter=. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 12:59, 2 February 2026 (UTC)

Are the various bugs listed in this section caused by newly added code, or are they bugs that have existed for a while and are only being reported now (presumably because of a sudden high number of them)? I ask because I am curious about the release and testing of new code. Is there a test page of citations that revised code runs against before being released? Or that could be used to agree on a desired outcome via discussion here? Mike Christie (talk - contribs - library) 13:26, 2 February 2026 (UTC)

The various bugs are most likely due to a lack of maintenance over the past ~2 years. The current maintainers do not seem to have the time to respond to and fix all the bugs. Anyone can help out on at: https://github.com/ms609/citation-bot should they be able to code in PHP. All proposed updates will/are still currently checked by @AManWithNoPlan

The test suite of the bot can be found at: https://github.com/ms609/citation-bot/tree/master/tests/phpunit/includes Redalert2fan (talk) 14:25, 2 February 2026 (UTC)

When the bot got blocked last time and I dived in to submitting patches to this repo, I was expecting it to be abandonware. Turns out the maintainer, AManWithNoPlan, is not very active on wiki but is extremely active on GitHub, constantly writing patches, approving patches (my patches all got merged within 1 day), and deploying the bot. So I would not say that this repo is inactive or that the maintainer is overwhelmed. I think it is well maintained.

The bot has an extremely big and thorough automated test suite. It takes over half an hour to run, running thousands of tests and testing external APIs. –Novem Linguae (talk) 19:07, 2 February 2026 (UTC)

The back end indeed is very well maintained, perhaps my message gave a wrong impression. I was intending to refer to bugs reported on this page.

And good news, the test suite no longer takes 30 minutes, it still does all the same tests as before but I managed to enable multiple process execution so it now only takes 7-8 minutes. Redalert2fan (talk) 19:20, 2 February 2026 (UTC)

Hehe no worries. I think it's normal for any complex software to have hundreds of open tickets, and for only the biggest bugs to get prioritized. –Novem Linguae (talk) 19:26, 2 February 2026 (UTC)

For the last few days before the block my watchlist was overrun by huge numbers of Citation bot edits. For the last day I can still see in its entirety on my watchlist, February 1, among still-visible changes on my watchlist, 90 out of 189 of the edits were made by Citation bot, mostly or entirely "Suggested by Abductive". The bot was blocked mid-day my time so the rate of changes was even higher on earlier days. It is unclear to me whether the greatly increased rate of problems was caused by changes to the bot code, changes to the underlying sources from which the bot draws its metadata, or just the extremely high rate of bot activity caused by Abductive's runs. —David Eppstein (talk) 18:45, 2 February 2026 (UTC)

Abductive's not the only one who used Citation bot. Some of us had to fix up some of the errors made after running that bot. sjones23 (talk - contributions) 18:49, 2 February 2026 (UTC)

That's why I said "still-visible changes". The 18 I linked at the top of this thread were from that day, for instance, but I didn't count them this time because they were no longer visible on my watchlist after edits by me and/or others. —David Eppstein (talk) 18:55, 2 February 2026 (UTC)

I figured that would be the case. sjones23 (talk - contributions) 18:56, 2 February 2026 (UTC)

Ok, I checked. 89 of the 90 were suggested by Abductive; one was by Headbomb (who I should say has always used the bot responsibly and not on the large-scale unsupervised fishing expeditions we tend to see from Abductive). —David Eppstein (talk) 19:01, 2 February 2026 (UTC)

"Some of us had to fix up some of the errors made after running that bot."

Yes, but because we unleash the article on specific articles, and we check after the bot, we catch and fix those as they come, and report issues as they arise. The occasional thing can slip by (as evidenced by DE's find above). You only have to look in the archives to find hundreds of my bug reports and tweak requests over the years. Abductive, OTOH, triggers the bot willy nilly on as many articles as possible and rarely if ever checks for errors. Headbomb {t · c · p · b} 21:49, 2 February 2026 (UTC)

Since the errors above all stem from {{citation}}, maybe the bot could temporarily stop touching {{citation}} while things are resolved? Headbomb {t · c · p · b} 21:50, 2 February 2026 (UTC)

That's not true of § issue and volume can't both be #163 and § Should not add useless nii.ac.jp and infoscience.epfl.ch urls above, at least. But there may be an observer effect here: I am reporting on things from my watchlist, which is disproportionately articles I have written, which disproportionately use Citation Style 2. The frequency of those issues among all articles may be very different. —David Eppstein (talk) 23:15, 2 February 2026 (UTC)

I mean there are other bugs, but they definitely disproportionally affect {{citation}}. The amount of bad edits I see accross articles using mainly CS1 style is well under 5%. And likely well under 1% too. Headbomb {t · c · p · b} 02:48, 3 February 2026 (UTC)

@Headbomb: In any case, the {{Citation}} part might need to be changed to match the appropriate online sources we've been using like {{Cite web}}, {{Cite news}}, {{Cite press release}}, {{Cite journal}} and so on. sjones23 (talk - contributions) 08:28, 12 February 2026 (UTC)

Some of this seems like really trivial bickering. For the first case above, Patricia Grambsch, does it really matter whether Mayo Clinic is publisher or work? If anything, the publisher is Mayo Foundation for Medical Education and Research and it doesn't seem that wrong to say the website is Mayo Clinic. Also the whole URL is a page not found. A little disputable annoyance about publisher vs work turns into a major issue because a particular editor is sending the bot off on thousands of articles and not checking anything. Is that really a constructive way to use the bot?  — Chris Capoccia 💬 01:00, 6 February 2026 (UTC)

Regarding the Mayo Clinic's website, it is a work published by Mayo Foundation for Medical Education and Research. sjones23 (talk - contributions) 08:12, 12 February 2026 (UTC)

I do think fully blocking the bot isn't super helpful when a lot of the issues are a bit semantic. Not being able to run the bot as it is, is a net-loss for the project. Lee Vilenski ^{(talk • contribs)} 21:15, 17 February 2026 (UTC)

Semantic issues are exactly the sort of decisions the bot should not be usurping from humans. And I think the bot is still available for individual edits, just not for the sorts of unsupervised and indiscriminate mass runs perpetrated by Abductive. —David Eppstein (talk) 22:15, 17 February 2026 (UTC)

If it is, it isn't where I'm clicking to use it. Run on a single page just says it is blocked. Lee Vilenski ^{(talk • contribs)} 22:55, 17 February 2026 (UTC)

You can invoke the bot yourself, see WP:UCB. Headbomb {t · c · p · b} 23:02, 17 February 2026 (UTC)

Oh, I usually use the web interface at on a single page which doesn't work. Lee Vilenski ^{(talk • contribs)} 23:04, 17 February 2026 (UTC)

After the previous unblock the max pages was set to 1 - see "Bot limited to only single page request". This is why you were able to do (only) a single page request upon till this new block (and not multiple pages and categories).

Now that the bot is blocked again using the bot via the interface is not possible at all, even for single pages, since it uses the bots account.

However, you can use the functions of the bot via the citation expander gadget (see WP:UCB). Edits will be made under your account, you will get an edit preview to see the changes that will be made on the page you use the gadget. Redalert2fan (talk) 12:23, 18 February 2026 (UTC)

I get "Error: Citations request failed" when I use the gadget. Stefen 𝕋ower ^{Huddle • Handiwerk} 00:02, 19 February 2026 (UTC)

Did you enable the gadget, went to the page you want to work on, press "edit" to go to the edit window and then press the "citations" button?

When using the gadget you will not go to outside of wikipedia, you stay on the page you want to edit. "Error: Citations request failed" is a message only displayed on https://citations.toolforge.org/ if you run the bot directly from there or when you use the "expand citations" from the toolbar, because that activates the bot to run on that page.

If the gadget doesn't work (or there is nothing to fix) you will just get an edit preview with "(No difference)". Redalert2fan (talk) 12:46, 19 February 2026 (UTC)

The gadget was already enabled, and I did the rest, and got the error message. Stefen 𝕋ower ^{Huddle • Handiwerk} 21:05, 19 February 2026 (UTC)

On a specific page or on multiple pages? Redalert2fan (talk) 10:36, 20 February 2026 (UTC)

I ran it on Hunter S. Thompson. I tend to give up when it fails the first time. 🙂 Stefen 𝕋ower ^{Huddle • Handiwerk} 20:52, 20 February 2026 (UTC)

Thanks for sharing the page, I see the error message as well now. That should not be happening, but is not related to the bot being blocked. Instead there is something on that specific page that causes the bot to fail and by extension the gadget as well. Redalert2fan (talk) 21:18, 20 February 2026 (UTC)

Thank you for reviewing that. Do you have any idea what that something could be so I can comb through the article for it? Stefen 𝕋ower ^{Huddle • Handiwerk} 21:28, 20 February 2026 (UTC)

I couldn't find anything quickly, but I've asked AManWithNoPlan if he can use debug on that page to see what is wrong. Redalert2fan (talk) 11:43, 21 February 2026 (UTC)

The file size (page size) was exceeded for the gadget. It has a lower limit than the bot itself to prevent time outs. The limit has been raised, so larger pages can be run with the gadget, but of course it is more likely that they will time out.

Why does the gadget time out more often on larger pages? Because webbrowsers will time you out, the bot does not have this issue. Redalert2fan (talk) 16:44, 21 February 2026 (UTC)

Thank you for the follow-up. I'll try this on some other articles soon. Stefen 𝕋ower ^{Huddle • Handiwerk} 20:19, 21 February 2026 (UTC)

so do I install the script to get this working on my own ...content editor find this very useful timesaver versus the trivial things that are being discussed that really relate to cosmetic edits. Moxy🍁 00:10, 19 February 2026 (UTC)

You can activate the gadget on your account or install the script, then you can use the "citations" button as explained above. That will make it so you get a preview of what the bot would propose to change on the page. Downside compared to the bot itself is that on pages with a lot of references it is slower and can time out, and that not all of the fixes that the bot can do will be made with the gadget. But in general it will work fine. Ofcourse, you are required to check the edits before publishing them. Redalert2fan (talk) 12:53, 19 February 2026 (UTC)

Actions taken to fix issue 5301

• Investigated all work-related issues tagged with #5301 and split up into 30 different reports (30+ citations)
• Discovered Trivy, PHPStan and PHP Phan tests of bot suite were broken. Implemented fixes: PR #5492 and PR #5494. Bot test suite now functions correctly again.
• Implemented broad fix for block reason: when publisher= is present work= is no longer added. PR #5497
• Strip en-dash and pipe taglines from Zotero for publisher/Work – will prevent “promotional” sections being added. PR #5497
• Added specific override for sportsworldi.com PR #5497
• Added drive.google.com to ZOTERO_AVOID_REGEX – “Google Drive” was already blocked as a title, but now metadata from Google Drive will no longer be used for any automatic citation additions (was already present for Google Docs). PR #5497
• Fix for other block reason: Implemented suppression for Zotero incorrectly parsing eatcs.org posting admin as "author" PR #5497
• Added warning in the web interface for when bot skips adding author= when publisher= is present, asking users to check if the current use of publisher= is correct. This should attract attention to manually change publisher= to work= where appropriate. PR #5497
• Temporarily added citations from all reports as tests to the internal bot test suite to verify processing after the broad fix. 29 passed, one failed (issue with the test itself, now fixed). Afterwards condensed the number of tests to those needed to safeguard the new changes. PR #5497
• Tested all full articles included in reports submitted on talk page using &ignore_block=1 via web interface (cannot write, only shows what output would be) - Passed
• Pasted individual citations in my sandbox and used gadget to see edit preview. Expected outcome: “no difference” (or only changes without work= ) - Passed

I think the bugs for Issue 5301 should be fixed now. If more needs to be done let me known, and otherwise if somebody else could format an unblock request that would be great. --Redalert2fan (talk) 18:02, 29 March 2026 (UTC)

This user's unblock request has been reviewed by an administrator, who accepted the request.

Citation bot (block log • active blocks • global blocks • contribs • deleted contribs • filter log • creation log • change block settings • unblock • checkuser (log) • SI)

---

Request reason:

Please see above. Redalert2fan has written a patch for the primary reason the bot was blocked and manually tested said patch, and AManWithNoPlan has merged and deployed it.

Accept reason:

See below.

–Novem Linguae (talk) 02:07, 30 March 2026 (UTC)

• @HJ Mitchell: - okay to unblock? PhilKnight (talk) 02:50, 30 March 2026 (UTC)

  I think @David Eppstein was the blocking admin. HJ Mitchell just turned off autoblock since it's a bot. –Novem Linguae (talk) 04:17, 30 March 2026 (UTC)

  Ok, unblocked. As I wrote in my unblock edit summary, I think the changes (and especially not adding work= when publisher= is present) go a long way toward resolving the problems on the technical side. As I also wrote, the problems are also behavioral as well as technical (editors abusing the bot by making large indiscriminate runs and not checking the results) but those can be handled on an individual-editor basis. —David Eppstein (talk) 05:29, 30 March 2026 (UTC)

  Should any further technical issues related to this block pop up I have plenty of time this week to adjust things. Otherwise I'll try to pick up as many of the other ones in the mean time. Redalert2fan (talk) 07:22, 30 March 2026 (UTC)

  {{fixed}}. (tag for archiving). Redalert2fan (talk) 14:57, 28 April 2026 (UTC)

More more more free DOIs

* 10.1002/lol2. (L&O Letters) full open access Headbomb {t · c · p · b} 07:17, 12 February 2026 (UTC)

There's also for Limnology and Oceanography and subjournals

• 10.1002/lno. (If they are older than 3 years, e.g. February 2023) L&O
• 10.1002/lob. (If they are older than 3 years, e.g. February 2023) L&O Bulletin
• 10.1002/lom3. (If they are older than 3 years, e.g. February 2023) L&O Methods

*10.4319/lo. (all older than 3 years old, so all are free access) L&O

• 10.1002/loe2. (all older than 3 years old, so all are free access) L&O e-Lectures
• 10.4319/lol. (all older than 3 years old, so all are free access) L&O e-lectures
• 10.1215/21573689 (all older than 3 years old, so all are free access) L&O Fluids
• 10.1215/21573698 (all older than 3 years old, so all are free access) L&O Fluids
• 10.4319/lom. (all older than 3 years old, so all are free access) L&O Methods

Headbomb {t · c · p · b} 07:41, 12 February 2026 (UTC)

Also,

• 10.21105 (Open Journal) Headbomb {t · c · p · b} 18:17, 12 February 2026 (UTC)

Just to confirm I see you put 10.1002/lol2. here but 10.1002/lol2 (without the .) in Module:Citation/CS1/Configuration/sandbox, which one is correct? --Redalert2fan (talk) 20:37, 12 February 2026 (UTC)

same for loe2 Redalert2fan (talk) 20:43, 12 February 2026 (UTC)

With dots. I fixed it. Headbomb {t · c · p · b} 21:27, 12 February 2026 (UTC)

In the module, the incipits of the doi suffix are treated as lua patterns (similar to regex). In lua patterns, like regex, a dot matches any character. If you really mean lol2. instead of loc2a, for example, then the lua pattern should be lol2%..

—Trappist the monk (talk) 22:40, 12 February 2026 (UTC)

Tweaked. Dots is what is meant. Feel free to correct. Headbomb {t · c · p · b} 23:17, 12 February 2026 (UTC)

And I'm personally not sure if there is support for "If they are older than 3 years" as in If the bot checks that so I'm just holding off on those for now. --Redalert2fan (talk) 20:37, 12 February 2026 (UTC)

Look for code around PNAS dois, with prefix 10.1073/pnas. There's a 6 month embargo on those, which I believe the bot handles (implemented around Aug 2021). Headbomb {t · c · p · b} 21:29, 12 February 2026 (UTC)

Nothing for special DOI handling comes up after a quick search for PNAS. Only specific handling for this is fixing capitalization.

10.1073/pnas is not in the free DOI list either. It is on a list of DOI's that need to be periodically checked by a human. It also is used in the a single test of the test suite for PMID and PMC expansion from a DOI. But all those things do not seem relevant to any embargo. Redalert2fan (talk) 00:13, 13 February 2026 (UTC)

Well then let's make that a feature request. Headbomb {t · c · p · b} 02:10, 13 February 2026 (UTC)

Would probably require extra external API interaction(s) Redalert2fan (talk) 11:57, 13 February 2026 (UTC)

Upon further investigation: 10.1073/pnas was hard coded in indeed before. Would make more sense to do it via a list instead for expandability and maintainability. Redalert2fan (talk) 16:50, 27 April 2026 (UTC)

Patch for DOI's without any time limitations: https://github.com/ms609/citation-bot/pull/5476. Put a strike trough in the text above to mark which ones are done. --Redalert2fan (talk) 12:44, 13 February 2026 (UTC)

Start with moving Free DOIs out of bad_data and give them their own location. PR 5505 Redalert2fan (talk) 20:02, 30 March 2026 (UTC)

Looking to implement 2 options: After year XXXX and older than XX Months.

After year XXXX example: DOI is free if published after 2014. If year/date/publication-date or API indicates something after 2014 will be tagged as free.

Older than XX Months example: If DOI is free after 3 years (36 months) and date/publication-date or API indicates that it is now older than 36 months because exact date is available tag as free. If only month is available tag at end of the month as free. If only year is available wait until end of year to tag as free, now worst case that would mean we would tag something as free 11 months late if only year is available, but would mean not prematurely tagging things as free, and is still better than no tagging at all.

Obviously for both, If no parseable date is present, no free tag is added. Redalert2fan (talk) 14:36, 28 April 2026 (UTC)

Included in PR 5539. Redalert2fan (talk) 15:33, 28 April 2026 (UTC)

Implemented and seems to be working: diff Redalert2fan (talk) 16:26, 29 April 2026 (UTC)

{{fixed}}. Redalert2fan (talk) 16:30, 29 April 2026 (UTC)

Tweak 10.1140/epjc is only free since 2014

Right now all 10.1140/epjc are marked free. This is my fault, only those after 2014 should automatically be tagged. I will be reviewing potential mistakes manually. Headbomb {t · c · p · b} 13:19, 29 March 2026 (UTC)

Understood, Automatic tagging depended on date is not implemented yet (as you requested earlier). I will remove 10.1140/epjc from the list until that is implemented. Redalert2fan (talk) 13:24, 29 March 2026 (UTC)

Included in PR 5539 Redalert2fan (talk) 15:33, 28 April 2026 (UTC)

{{Fixed}}. Example: diff Redalert2fan (talk) 16:29, 29 April 2026 (UTC)

Cleanup Visuel Editor crap issue for Annual Reviews (10.1146/...)

This affects 3-4 citations a day, and is a pain to cleanup by hand. Headbomb {t · c · p · b} 00:37, 15 April 2026 (UTC)

PR 5543 --Redalert2fan (talk) 21:01, 28 April 2026 (UTC)

The fix for this got accidentally overwritten in a merge, but has been submitted again. Awaiting implementation once more. Redalert2fan (talk) 16:31, 29 April 2026 (UTC)

@Headbomb After code implementation result is: diff. Redalert2fan (talk) 20:24, 29 April 2026 (UTC)

created broken citation template: |vauthors=

Patch PR 5544. --Redalert2fan (talk) 21:40, 28 April 2026 (UTC)

Now correctly recognises 3rd, but needs some further tweaking to prevent it being added as a seperate last= or first= incorrectly. Redalert2fan (talk) 08:01, 29 April 2026 (UTC)

New patch: PR 5546. Redalert2fan (talk) 16:19, 29 April 2026 (UTC)

And the new result: diff. Redalert2fan (talk) 16:21, 29 April 2026 (UTC)

Ummm, not really. This diff is different from the original diff. In the original diff, the bot created a template that uses the |vauthors= parameter whereas in your new diff, the bot created a template that uses |lastn= / |firstn=.

According to doi:10.1007/978-3-540-69248-5_2 the author's name is Peyton Jacob III. There is no need to convert III to 3rd when including the generational suffix in |first=. In fact, such a conversion is discouraged and you will notice that the converted parameter (|first3=P. 3rd) causes Module:Citation/CS1 to emit a CS1 maint: numeric names: authors list maintenance message.

When creating |vauthors= or |veditors= parameters, generational suffixen (with the exception of Jr) must be numeric ordinal numbers. When creating |firstn=, generational suffixen may be text or numeric; if numeric then the whole parameter value should be wrapped in ((...)) accept-as-written markup to suppress the maintenance message and category.

—Trappist the monk (talk) 17:23, 29 April 2026 (UTC)

The Pubmed PMC API returns it as <Item Name="Author" Type="String">Jacob P 3rd</Item>, the bot does not convert III to 3rd itself. The API used if you query doi:10.1007/978-3-540-69248-5_2 will return Peyton Jacob, without any suffix. --Redalert2fan (talk) 18:30, 29 April 2026 (UTC)

I also don't see the CS1 maint: numeric names: authors list maintenance message. Or are those disabled in my userspace sandbox? Redalert2fan (talk) 18:40, 29 April 2026 (UTC)

(edit conflict)

See Help:CS1 errors § Controlling error message display.

—Trappist the monk (talk) 18:50, 29 April 2026 (UTC)

Thanks, I've enabled them now. Redalert2fan (talk) 19:12, 29 April 2026 (UTC)

(edit conflict)

Consistency is wonderful thing, isn't it? Would that Springer were internally consistent. They offer recommended citations in various formats at https://link.springer.com/chapter/10.1007/978-3-540-69248-5_2#citeas; none of which include the generational suffix. Yet, at the top of the doi page the author's name is listed as Peyton Jacob III. For that name, the Vancouver format favored by PMC and PMID, Jacob P 3rd, is correct and is acceptable to cs1|2 when used in |vauthors=.

Still, the object of this bug report was the creation of a |vauthors= parameter with a malformed author name. That problem has not been demonstrated to have been fixed. Has it been fixed?

—Trappist the monk (talk) 18:50, 29 April 2026 (UTC)

After the code change the bot no longer creates |vauthors= for this citation and chooses the other route instead. Since I hadn't got the error messages enabled and the new result included "P 3rd" I assumed it was fixed because of that. Redalert2fan (talk) 19:12, 29 April 2026 (UTC)

@Trappist the monk Ah I was slightly incorrect, I overlooked something. the original page has cs1 config|name-list-style=vanc forcing citation bot to use |vauthors=, while in my sandbox it was free to use |lastn= / |firstn=. So it was not the code change.

This is what the updated code gives: diff. So the reported bug was actually fixed.

However I guess we just discovered that the bot can also introduce CS1 maint: numeric names: authors list errors. So ((...)) wrapping should be introduced per your text above. Redalert2fan (talk) 20:04, 29 April 2026 (UTC)

Patch for wrapping: PR 5550. Redalert2fan (talk) 20:33, 29 April 2026 (UTC)

Wrapping: diff

@Trappist the monk With the updated code diff above and this diff please let me know if you consider the issue resolved. Redalert2fan (talk) 14:03, 30 April 2026 (UTC)

Yeah, I think so – as long as the bot doesn't attempt to change this (created by the abomination that is visual editor):

|last=April 22 |first=Jordan Willow Evans on |last2=Pm |first2=2026 3:41

to this:

|last=((April 22)) |first=Jordan Willow Evans on |last2=Pm |first2=((2026 3:41))

The accept-as-written markup would prevent Module:Citation/CS1 from emitting the proper maintenance message and category which would hide the corruption from editors who might apply a remedy.

—Trappist the monk (talk) 14:25, 30 April 2026 (UTC)

It won't: diff. Redalert2fan (talk) 14:55, 30 April 2026 (UTC)

Implementing Web Content Accessibility Guidelines

I will be starting implementing the Web Content Accessibility Guidelines starting with version 2.1 to comply with at least WCAG 2.1 Level AA. Currently the web interface does not comply with most of these standards so it might be difficult to use the web interface for some. Implementing this will improve the Accessibility of the bot's web interface for users with certain disabilities. Examples could currently be (but not limited to) users with colour blindness who are not able to distinguish fields from background / see changes made, or users using screen readers experiencing bugs with their text to speech not able to correctly interact with the page. Redalert2fan (talk) 16:25, 28 April 2026 (UTC)

Support Seems to be a good thing. Grimes2 16:35, 28 April 2026 (UTC)

Groundwork implementation: PR 5540. Redalert2fan (talk) 22:11, 28 April 2026 (UTC)

WCAG 2.1 Level AA: PR 5545. Redalert2fan (talk) 08:52, 29 April 2026 (UTC)

Most of it has been implemented. Redalert2fan (talk) 20:18, 30 April 2026 (UTC)

Currently looking in to optionally implementing 2.2 Level AA (not mandatory). Redalert2fan (talk) 15:42, 4 May 2026 (UTC)

Implemented a few items from 2.2. Should anyone have any trouble with any other aspects accessibility of the bot's web interface please feel free to start a new topic. Marking this as {{fixed}}. Redalert2fan (talk) 12:47, 9 May 2026 (UTC)

Promotional title

Thank you for your report. I see the edit you shared is from December 2022 and I am unable to reproduce this addition right now. Code changes in the past years likely have resolved this. --Redalert2fan (talk) 13:02, 9 May 2026 (UTC)

10.1002/cesm / Cochrane Evidence Synthesis and Methods is open access

Patch: PR 5573. --Redalert2fan (talk) 16:05, 12 May 2026 (UTC)