Wikipedia talk:Reliable sources

I am requesting input on where the community stands on the idea of GA and FA reviewers asking for page numbers in journal articles (by this I mean the relevant pages within the page rage of the journal article, i.e., {{sfn|Smith|1990|pp=90-91}}), similar to books to improve the verification process during the course of a review even if it is only for 5-10% of the sources in question. While it is easier to pin point the relevant pages in history and political science topics, which is effectively a standard in these WP articles, in natural sciences and math articles follow the academic tradition of not providing such information. The reaction to this idea was surprising hostile in another discussion, despite IMO the obvious benefits coming from this. Nevertheless, the arguments presented in a previous discussion were:

1. It is good for the reader/reviewer to read all the pages of the journal paper, even if there are 50+ such sources in the WP article
2. Citation points to the whole journal article and so it is pointless to provide page numbers

Both these arguments are problematic as they run afoul to the WP policies.

1. WP is not an academic journal (thus we should not use the reference style of academia, i.e., not providing pages) and it is not here to promote any of the sources used in verification WP:PROMO
2. If the relevant information is not found within the journal article, i.e., no specific page(s), one has to draw the conclusion from the whole paper then there is the danger of WP:SYTH and WP:OR
3. More important than most is that easing the review process the reviewer will have more time to evaluate more of the presented sources and perhaps provide better recommendations for the improvement of the article.

While these concerns will be pushed aside, because who care what I say. However, things have changed in academia as it produces fake papers in record numbers:

• Fraudulent Scientific Papers Are Rapidly Increasing, Study Finds New York Times
• The situation has become appalling’: fake scientific papers push research credibility to crisis point Guardian
• Fake scientific papers are alarmingly common by Science
• Scientific fraud has become an ‘industry,’ alarming analysis finds by Science
• How big is science's fake-paper problem? by Nature
• Fake studies in academic journals may be more common than previously thought by NPR
  • Well, the goal of our study was really to estimate the scope rather than identifying a specific paper - this is fake. So when we're looking at the numbers of papers that are identified with our indicators, the percentage was in 2020 at 28% of all biomedical publications. That comes to over 300,000 in the biomedical field alone. Now, if you consider that all of science is maybe roughly double that, then you can sort of roughly estimate that there may be a half million fake papers published per year. And that is a shocking number. (added emphasis)

You can shrug by saying this does not affects us, right? Wrong! Many of these bad papers end up on WP pages even after being retracted and there is already a study of how bad we are doing:

• The Persistence of Retracted Papers on Wikipedia
  • Our findings highlight how the Wikipedia community supports collaborative maintenance but leaves gaps in citation-level repair. ... Our findings reveal that while some citations are corrected promptly, many persist uncorrected for extended periods, with a median time to correction of 3.68 years (1,344 days). ... Conversely, a high academic citation count is associated with a slower correction time, which may reflect the difficulty of challenging sources perceived as highly authoritative. ... These findings highlight a fundamental disconnect between the public availability of retraction data and the community’s collaborative workflows, which often fail to translate this information into the reader-facing edits required for a complete repair. (emphasis added)

Irrespective of your opinion on the subject matter, the key issue is that it has been recognized as a problem in the "community's collaborative workflows" and I do not think inaction is an option. My opinion is based that if we more easily verify then we have more time to evaluate more sources and better appreciate which of the sources are good and which are not. Maybe my opinion was wrong, so I want to hear from the community how this could be addressed. PS: keep it WP:CIVIL Thank you. A.Cython(talk) 04:03, 28 February 2026 (UTC)

Could you explain the relevance of fake papers and retractions to your desire for more page numbers? I would agree that fake papers and retractions are both problematic, but I don't see how having specific page numbers or not would possibly address that. Nikkimaria (talk) 05:16, 28 February 2026 (UTC)

If an article under review has (as an example) 50 journal articles with on average of 20 pages, then the number of pages a reviewer has to go through is 1000 pages. This is a bit excessive, but what it means is that the reviewer will spend way to much time just reading these 1000 pages in order to spot how the citations of these journal articles support the relevant statements in the article. I mean we are human and we can volunteer only so much time so the reviewer's time is valuable. As a result the number of journal articles that will be evaluated will be small. If on the other hand, we ease the review process even by a bit, then the reviewer will have more time to evaluate more of these journal articles and perhaps being in a better position to spot one of these problematic papers. It is like the joke "If you have one watch you know the time, but if you have two you do not", except here to spot the watch that is broken one needs to have more watches. As I said I might be wrong (and let's assume that I am). So I really want to hear others on how to approach this issue. Do we have alternative ideas? A.Cython(talk) 05:50, 28 February 2026 (UTC)

I believe that the accepted practice for Wikipedia:Good articles is to spot-check 10% of the sources, which means that 50 articles × 20 pages × 10% = 100 pages.

Also, 6–10 pages is more likely, so it's really 30–50 pages.

And you don't usually have to read the whole thing to determine that the source supports the claim, so the math is frequently closer to 10% of 50 articles @ average of 2 pages each = 10 pages total. WhatamIdoing (talk) 06:06, 28 February 2026 (UTC)

That's my point! As the manuscript claims, we end up with errors! So either your math has the problem (no offense) or something else needs to change. If within the journal it is only two pages that are of interest why do I need to waste time to find them in the as you say 10 page paper? Would it not more useful to have a little help to narrow down the search so that a reviewer can read more than 10% and perhaps being able to spot the bad ones? A.Cython(talk) 06:16, 28 February 2026 (UTC)

On what percentage of (~scientific) articles are you finding that you actually need more than a couple of minutes to check the source? Perles configuration is about advanced mathematics, so that's going to be difficult for a lot of reviewers, but thinking generally about articles in the hard sciences, do you find that you're basically at sea in all the sources, or do you often find yourself thinking "It says 39.2% of foo, so ⌘F – and there it is: 39.2% of foo is confirmed"? WhatamIdoing (talk) 07:17, 28 February 2026 (UTC)

I cannot give you a percentage, but articles in natural sciences and maths tend to have much more sources from academic journals without page numbers or other locator to narrow down than any other field, e.g., history and political science (I do not include in the discussion Video Games, Music, etc). This is natural since most people who add information in these articles are coming from an academic background and thus they do what they have been trained to do (i.e., no page numbers in citations). This is not against any WP policy, but the question that I raise is how do we organize this information in order to be more efficient later on. Should we treat research articles as WP:PRIMARY sources, but academic reviews as WP: SECONDARY? Should we add page numbers to ease verification? I do not know, which is why I would appreciate to hear potential solutions rather than hearing how wrong I am.

• Strangely enough, the arguments raised in previous discussion (listed above) could have easily raised/applied in history and political science articles, but my understanding this is not the case. I suspect that the editors/reviewers need to find the information more quickly to properly evaluate them, since these topics sometimes can be contentious. Having time to evaluate this is critical for a good review.
• In my limited experience (and take this with a grain of salt) new editors (probably more familiar with WP in their life) are more willing to adopt the inclusion of page numbers than more experienced editors. So time is the solution?
• advanced mathematics, so that's going to be difficult for a lot of reviewers No, the issue was not that the topic is difficult, but the response to the reviewer. In another turbulent review, we had to go through conspiratorial accusations (I am tanking the review for GA Backlog review points), lecturing, etc, to eventually work it out for a handful inclusions. Given that the similar attitude is expressed by other editors, i.e., the reviewer is extremely lazy or deluded, it appears to be a people-problem and not content-problem. Moreover, the refusal to even contemplate change only amplifies in my mind and potentially others the fundamental disconnect ... the community's collaborative workflows, i.e., if editors interpret the current policy as a way to reject reviewer's recommendations then what is the point of the reviewer in the first place? As I said, let's assume that I am wrong, then what is a potential solution?

It is clear that we will stay in status quo and wait the next academic paper saying how bad we are doing. Happy editing! A.Cython(talk) 17:52, 28 February 2026 (UTC)

I don't think this logic works, in terms of the retractions/false papers piece. If we assume that reviewers are reading the entire article when not provided with specific page numbers, that would make them more likely to spot problems with the piece as a whole, not less. If we assume they are only spotchecking, it shouldn't make any difference.

In terms of alternative ideas... suggest people install PubPeer or a similar tool? Nikkimaria (talk) 14:57, 28 February 2026 (UTC)

Thanks for the tool, but to be used efficiently the reviewer needs to have time. Page numbers at least in my mind helps to reduce the time spend in one task and redistribute this valuable commodity on other aspects of the review such as evaluating the quality of the sources or perhaps evaluating more than 10% of the sources. Staying at 10%, as the manuscript mentioned above demonstrated, is insufficient. A.Cython(talk) 17:58, 28 February 2026 (UTC)

Assuming that there is a set time per review and that reducing the time spent on one aspect will necessarily result in additional time being spent on what you consider more valuable. I don't think either of those are safe assumptions. If (as WAID suggests) the practice is 10%, then the practice is 10%, and that's the practice you'd want to change. Nikkimaria (talk) 01:56, 1 March 2026 (UTC)

Thank you for this insightful comment. A.Cython(talk) 02:05, 1 March 2026 (UTC)

See also Wikipedia talk:Good article nominations#References in a GAN article. Maybe we should just have one discussion per the WP:SEETALK guideline, to avoid a WP:TALKFORK? WhatamIdoing (talk) 05:58, 28 February 2026 (UTC)

I was told to bring the discussion here. Given that the previous discussion started with the wrong foot, I felt prudent to write a better intro by summarizing the various constructive comments before insults like the reviewer is extremely lazy etc start flying around so that we reach to a useful outcome. A.Cython(talk) 06:03, 28 February 2026 (UTC)

• How is this a WP:RS issue? Despite the discussion above, I still see no connection between page numbers and retracted or fake papers (if a paper is not a WP:RS due to issues with publication, providing page numbers won't help us recognize that) and neither are really RS issues (RS is not and has never based on the idea that any sort of source is perfect; even the best of sources can make errors, what matters is how they handle them.) Regarding the issue with retracted papers we could perhaps run some sort of bot to find and tag or list citations to papers that have been retracted, but even that isn't really an RS issue in the sense that there's no real disagreement about reliability itself; it's a technical issue in terms of how we can better catch them. --Aquillion (talk) 17:24, 6 March 2026 (UTC)

  User:RetractionBot does this. WhatamIdoing (talk) 18:32, 6 March 2026 (UTC)

  For reference RetractionBot adds articles to the following categories:
  Category:Articles citing retracted publications
  Category:Articles citing publications with expressions of concern
  Category:Articles citing publications with errata -- LCU ActivelyDisinterested «@» °∆t° 19:27, 6 March 2026 (UTC)

These two subpages seem to use "usurped" to mean two distinct things, which is a bit confusing to me:

• Wikipedia:Link rot/Usurpations (WP:USURPURL) says "Usurped domains are used by spammers, squatters, malware, SEO, phishing or other fraudulent activity. Typically they are legitimate domains that expired and were hijacked."
• Wikipedia:Reliable sources/Usurped sources (WP:USURPSOURCE) says "Usurped sources are websites (NYT, Guardian, etc.) whose content has been copied to another website, without attribution or faking who wrote it."

I think of "usurped" as having the first meaning because of the url-status=usurped attribute in templates like Template:Cite web and Template:Usurped. The second page is helpful but could be renamed something like "Plagiarized sources"? Also, both subpages would benefit from hatnotes indicating whether they are essays, information pages, etc.

Courtesy ping to @GreenC, who seems to have been the primary author for both. :) Dreamyshade (talk) 03:18, 1 March 2026 (UTC)

The word "usurped" has multiple uses on Wikipedia none have priority. WP:USURP dab page. The concept of USURPSOURCE was not invented by me, but the name did come about through mutual discussion with at least User:Grayfell maybe others. Plagiarism doesn't adequately convey the scale and bad faith element, these are mass thefts often for purposes like disinformation and malware, it is more akin to usurping entire websites vs. the lighter ethical concern of not giving an author due credit. -- GreenC 05:15, 1 March 2026 (UTC)

Thanks for the context! Might help to simply put in more specific disambiguation hatnotes; I'll give that a try. Dreamyshade (talk) 16:27, 1 March 2026 (UTC)

Quentin was a man. Why is he being referred to as ‘She’? ~2026-13369-28 (talk) 13:42, 1 March 2026 (UTC)

The place to ask such questions would be the articles talk page (talk:Quentin Crisp), but it looks like the explanation is in the Quentin Crisp#Gender identity section. -- LCU ActivelyDisinterested «@» °∆t° 16:36, 1 March 2026 (UTC)

Where would think tanks fit in on this? We should create a section for those with a shortcut WP:THINKTANKSRS or WP:THINKTANK Iljhgtn (they/them · talk) 16:02, 6 March 2026 (UTC)

The answer is "it's complicated". WhatamIdoing (talk) 18:32, 6 March 2026 (UTC)

My guess is that there has to be an RFC to put an entry into the main WP:RSP table. Given the current events, it looks like Iran International is going to be asked about often, especially since it's heavily used in Wikipedia. For the moment we have Wikipedia:Reliable sources/Noticeboard/Archive 506#Is Iran International a reliable source?, without a formal summary. I'm not going to propose an RFC, but that might be useful if there are frequent requests. Boud (talk) 10:29, 7 March 2026 (UTC)

Iff there are multiple requests, then an RFC would be typical per Wikipedia:Reliable sources/Perennial sources#Inclusion criteria. But it's also possible that editors will settle into the ordinary pattern: "Is this a news outlet? Yes, but they're in a war, so anything they say about the war should be considered WP:RSBIASED (ditto for the other side)". In that case, while the source might get asked about, the discussions might get resolved quickly, too, without needing to add anything to RSP. WhatamIdoing (talk) 15:48, 7 March 2026 (UTC)

What guidance is there on using museum websites and the display boards within museums? In asking this question, I have in mind the following:

• Some museums are run by noted academics and the quality of their output is high. e.g. Vasa Museum, under director of research Fred Hocker (strange that there is no article on him).
• Murray Pittock's criticism: "the National Trust for Scotland (NTS) at the [Culloden] Visitor Centre still describe Elcho’s and Balmerino’s Life Guards, Bagot’s Hussars, and Strathallan’s Perthshire Horse as ‘Highland Horse’, though there was nothing ‘Highland’ about any of them." (Pittock, Murray. Culloden: Great Battles (p. 135).} I recently checked with the visitor centre and they maintain that their labelling is correct and have no regard for the leading historian of the subject calling them out in his 2016 book.
• This website which, though I can't prove it, really looks to me as though it has sourced its content from Wikipedia, taking some of the worst of our hyperbole and building on it a little.
• Personal experience of working with a local visitor attraction with ambitions to be a museum: the "museum experts" that they had hired seemed to have no understanding of the concept of recording where display board information came from (for internal record-keeping purposes) and seemed to think anything in print was a good source.

ThoughtIdRetired _TIR 21:43, 9 March 2026 (UTC)

You can use {{cite sign}} for the display placards. Generally, they are accepted for uncontroversial claims.

Fundamentally, a reliable source is one that editors are willing to rely on for the specific claim in question, we trust you to make a good decision about whether to rely on any given webpage/sign/whatever. WhatamIdoing (talk) 03:11, 10 March 2026 (UTC)

Whilst appreciative of the answer, this seems to be a bit of a circular argument. How do you know that a claim is uncontroversial without an RS to guide you? I think the best an editor can do is examine other content at the museum and see if that is accurate when compared with RSs. Then if that test is passed, use the museum information if there is no other RS to support the fact in question. I wonder whether some sort of warning of the variable quality of museum information is needed? ThoughtIdRetired _TIR 08:46, 10 March 2026 (UTC)

You assuming that a claim is uncontroversial when there are no sources disputing it. That's the same as any claim in any source. Uncontroversial does not mean that it is guaranteed to be correct. It just means that there is no existing, known controversy about it. WhatamIdoing (talk) 17:48, 10 March 2026 (UTC)

...To say nothing of normally-highly-respected institutions that may be pressured to revise history to serve a political narrative or a thin-skinned leader.

I suspect that challenges to the reliability of museum plaques come up rarely enough that we would tend to address them on a case-by-case basis. Specific concerns should be addressed at Wikipedia:Reliable sources/Noticeboard.

To address the more general case, I was going to point you to Wikipedia talk:Reliable sources, but it looks like this discussion is already duplicated there...? (Nevermind this bit. For some reason I could have sworn I saw this discussion posted on the Ref Desk. What a strange hallucination on my part. TenOfAllTrades(talk) 13:09, 10 March 2026 (UTC)) I'm not sure that museum signage reliability is a frequent-enough concern to require a special callout on the main WP:RS policy. TenOfAllTrades(talk) 11:48, 10 March 2026 (UTC)

I think we get a question here/on WP:PG talk pages about every other year. WhatamIdoing (talk) 17:49, 10 March 2026 (UTC)

• Though generally reliable, their reliability varies considerably with the size, status and budget of the museum concerned, and how recently they have been updated. For example, locally-run Italian museums (as opposed to the bigger nationally-run ones) can be rather iffy (and their English translations poor, which is a problem even in the Louvre). Ideally other sources from the same museum should be used. Johnbod (talk) 18:58, 10 March 2026 (UTC)

Mike Caulfield developed the SIFT method for determining the reliability of a news source. However, I haven't seen it discussed here or on Wikipedia:Reliable sources/Noticeboard. Is it generally considered a useful method by Wikipedia editors? Ixfd64 (talk) 21:01, 16 March 2026 (UTC)

Just skimming the first part of the Investigate the Source says "Would their assessment be biased?", but Wikipedia explicitly allows for biased sources and that bias isn't a relevant factor in reliability (WP:RSBIAS). As with many external sources that deal with source reliability, it might be useful for individual editor and their investigations into a source, but it isn't based on Wikipedia's policies and guidelines. Although there are many discussions on 'reliability' the actual point is 'reliability for Wikipedia's purposes' not some general understanding of reliability without that context.
So useful for getting editors to think about how to look at sources, but any assesment just based on SIFT may not include the complete context. -- LCU ActivelyDisinterested «@» °∆t° 21:30, 16 March 2026 (UTC)

I wonder if it could be mentioned in an article about Media bias. For editors in particular, someone could always write an essay (either about that one in particular, or about multiple). WhatamIdoing (talk) 23:57, 16 March 2026 (UTC)

I had not heard about it before. After reading it, I find it both vague and subjective, and doubt if it will be widely used on Wiki. And as Actively said, his concept of reliable is not based on Wiki policy. Sorry, but for yours truly it is a no go. Yesterday, all my dreams... (talk) 10:11, 17 March 2026 (UTC)

I definitely get the sense that it's not thorough enough for Wikipedia standards. Ixfd64 (talk) 17:23, 18 March 2026 (UTC)

You are right. Yesterday, all my dreams... (talk) 21:51, 18 March 2026 (UTC)

It could be useful in educating editors on how to look at sources when assessing reliability, but it's not a complete picture. -- LCU ActivelyDisinterested «@» °∆t° 12:49, 19 March 2026 (UTC)

I'm thinking about writing an essay on how discussions from foreign-language Wikipedias affect source reliability on English Wikipedia. Specifically, if a source is discussed on, say, the Spanish Wikipedia and deemed reliable there, should we treat it as reliable by default unless we've discussed it ourselves? How would that work? Gommeh ^{(talk! sign!)} 14:53, 18 March 2026 (UTC)

The different language version run entirely separately, so whatever is said on one has no barring on a different language version. Even policies and guidelines aren't the exact same across different version, so judgements on one could be based on ideas that don't work on another. -- LCU ActivelyDisinterested «@» °∆t° 15:19, 18 March 2026 (UTC)

Apart from what Actively said, remember that in places like China usres are very careful about what they say, if they know what is good for them. So while links from those my be useful reliability is not transferred. Yesterday, all my dreams... (talk) 08:14, 19 March 2026 (UTC)