Talk:Hallucination (artificial intelligence)

Merging with Hallucination (artificial intellligence)

It has been suggested that the Hallucination (NLP) article should be merged with this article. I agree with this proposal. The reason I created the Hallucination (NLP) article is because I did not find this one: it is not referenced anywhere, even in the Hallucination (disambiguation) page. However I think that the Hallucination (NLP) article has some valuable new content which is properly sourced and not in this article, so there is value to add the current content of the Hallucination (NLP) article. Hervegirod (talk) 17:47, 16 January 2023 (UTC)

I just did it, seems uncontroversial and basically the same topic. And because 'artificial intellligence' is broader than NLP, it makes sense to merge into a broader article. Artem.G (talk) 18:41, 30 January 2023 (UTC)

Some Examples of Hallucination are better than others.

The article currently states: Mike Pearl of Mashable tested ChatGPT with multiple questions. In one example, he asked the model for "the largest country in Central America that isn't Mexico". ChatGPT responded with Guatemala, when the answer is instead Nicaragua. While this does at first seem to be an error on ChatGPT's part, consider that Pearl did not specify "Second largest by land area", of which Nicaragua is. It is entirely possible that ChatGPT interpreted his question as referring to population, in which case Guatemala is the correct answer. There is not enough information provided here to determine whether or not this was genuine AI hallucination or just GPT misinterpreting the vague data it was asked for. 192.77.12.11 (talk) 05:23, 15 March 2023 (UTC)

It's an error either way ("largest" almost always means "largest by area" rather than "most populous"), but I removed it since there are plenty of other examples. Rolf H Nelson (talk) 18:12, 19 March 2023 (UTC)

Totally disagree. Google "world's largest democracy", and then compare the top result with the one that is the largest by land area. Mathglot (talk) 10:59, 29 March 2023 (UTC)

also by now, even though I know it's been several years since this comment, it would be fair to point out that this model that was used for this question is now considerably old, outdated, and officially decommissioned.

Trying this question now with GPT 5.6 Sol would be appropriate. The article is heavily out of date and I'm working on it PlutoidRepublic (talk) 19:15, 31 July 2026 (UTC)

Hallucinating non-existent APIs

A German geocoding company was flooded by dissatisfied customers trying to use code ChatGPT wrote: https://the-decoder.com/company-wins-customers-via-chatgpt-for-a-product-it-does-not-carry And EleutherAI complained people keep trying to access a URL they don't have on their website: https://twitter.com/AiEleuther/status/1633971388317941763 Likely, there should be more examples, which ones are notable? Ain92 (talk) 10:05, 16 March 2023 (UTC)

I didn't find any great WP:RS searching for (eleuther hallucination) nor (opencage hallucination), so per WP:SYNTH I'm pesonally reluctant to add either until we get a strong reporting source explicitly calling one of them a hallucination. Rolf H Nelson (talk) 18:20, 19 March 2023 (UTC)

'confidence' - is this a rigorous term?

The introduction defines a hallucination as "a confident response". In this context, is 'confidence' being used as statistical concept (e.g. confidence interval) or does it just mean that the generated text reads as if the writer were confident? If this 'confidence' is based purely on the text, I think the hallucination should be described as "seemingly confident", because there is no underlying assessment of confidence by the machine. AdamChrisR (talk) 12:51, 8 April 2023 (UTC)

@AdamChrisR the phrasing "a confident response" is terse, reflects the sources, and seems accurate to me, even under (say) "mimicry" models. I can say "Harry Potter was confident that his life would be peaceful" even though neither Daniel Radcliffe nor J.K. Rowling nor any concrete entity made such an assessment of confidence. That said, if we can find a strong source for alternate views, we should include them. Rolf H Nelson (talk) 02:37, 11 June 2023 (UTC)

Intro is way too wordy

"Note that while a human hallucination is a percept by a human that cannot sensibly be associated with the portion of the external world that the human is currently directly observing with sense organs, an AI hallucination is instead a confident response by an AI that cannot be grounded in any of its training data."

That sentence clearly had a lot of work put into it so I didn't touch it, but imho, it should be much simpler. Something like:

"While a human hallucination is when a person sees or feels something that doesn't match up with what's actually happening around them, an AI hallucination is instead a confident response by an AI that cannot be grounded in any of its training data." — Preceding unsigned comment added by Cainxinth (talk • contribs) 21:01, 10 April 2023 (UTC)

Not only is the intro too wordy, it is also incorrect. Where exactly is the proof/source that such confident responses "are not grounded in any of its training data" ? If an AI chatbot was trained on fruits, then surely contaminated or corrupted data could end up provide responses that don't involve fruits. But that doesn't make it "not grounded in any training data". --AloisIrlmaier (talk) 14:37, 26 April 2023 (UTC)

I agree that it's an incorrect definition. Using the example from The Internal State of an LLM Knows When its Lying, an LLM may decide that the most likely word to follow "Pluto is the" is "smallest", but then have no high-probability completions (the paper suggests "dwarf planet in our solar system" and "celestial body in the solar system that has ever been classified as a planet.", both of which are incorrect). It's like the LLM 'knows' that none of its suggestions are accurate, yet it's painted itself into a corner. So "not grounded in any training data" doesn't seem accurate here, either. (Possibly this specific type of occurrence would be better handled with a beam search, but that still only gives you local planning.) - CRGreathouse (t | c) 18:21, 5 May 2023 (UTC)

The definition of "Hallucination" (AI) needs to be reconsidered

The definition of the term is blatantly incorrect and lacks sources. Where exactly is the proof/source that such confident responses "are not grounded in any of its training data" ? If an AI chatbot was trained on fruits, then surely contaminated or corrupted data could end up providing responses that don't involve fruits. But that doesn't make it "not grounded in any training data". It seems like that the authors of this article are trying to dogwhistle that hallucination in AI chatbots might involve emergent behavior. This couldn't be any further from the truth. --AloisIrlmaier (talk) 14:40, 26 April 2023 (UTC)

@AloisIrlmaier @User:CRGreathouse The definition was sourced to "Survey of Hallucination in Natural Language Generation" but similar definitions appear in other contexts. Is there an alternative WP:RS that you (or others) would prefer? Rolf H Nelson (talk) 00:41, 11 June 2023 (UTC)

I don’t have an alternate source to suggest offhand, but I agree that this definition is bad. I strongly agree with you that renaming the page is inappropriate as hallucination is overwhelmingly the term used. (Confabulation has its own issues, though it may be an improvement, but that’s not up to us to decide but the broader scientific community.) - CRGreathouse (t | c) 01:08, 12 June 2023 (UTC)

I agree with OP and would like to suggest renaming the page to Confabulation (AI). Confabulation more accurately represents what is described on this page. People who have a _stake_ in anthropomorphizing AI for their own benefit because anthropomorphizing it makes it more engaging, and therefore economically valuable, use words like hallucination _strategically_. It's not objective. An objective description would be confabulation. As AloisIrlmaier is getting at, the definition of hallucination is an internal experience not grounded in reality. AIs don't have an internal experience and certainly not one that isn't grounded in reality. Confabulation is defined by visible behaviors based on fabrication, which is exactly what is happening here. TwigsCogito (talk) 12:19, 11 June 2023 (UTC)

Renaming is a non-starter at this time, all the sources acknowledge that the current mainstream terminology is "hallucination". You need to lobby the scientists and the mainstream media, not Wikipedia, if you want to change that. Rolf H Nelson (talk) 19:23, 11 June 2023 (UTC)

there is a page for it Confabulation (neural networks). Artem.G (talk) 04:45, 12 June 2023 (UTC)

I feel that delusion is the proper term. I also feel that this page carries enough weight that discussing the difference between "an experience involving the apparent perception of something not present" and "a false belief or judgment about external reality, held despite incontrovertible evidence to the contrary, occurring especially in mental conditions" will at least allow folks to question the current thinking. 216.213.180.191 (talk) 19:28, 3 September 2023 (UTC)

A definition of hallucination for IA was added in September 2023 to the Merriam-Webster dictionnary (https://www.merriam-webster.com/wordplay/new-words-in-the-dictionary). Their definition is : a plausible but false or misleading response generated by an artificial intelligence algorithm. Bob20230408 (talk) 08:59, 28 October 2023 (UTC)

Who coined usage of “hallucination” with respect to AI models

I don't know the answer myself, nor am I sure where to find it or source it, but I think it would be very interesting if the article could tell who coined the usage of "hallucination" for referring to AI model output, and when. Showeropera (talk) 16:25, 13 July 2023 (UTC)

The reference to WOPR “hallucinating” stems from pop culture, and it’s unclear whether that portrayal reflected artificial intelligence or simply a malfunctioning computer program. In contrast, Thaler’s work focused on uncovering the mechanisms by which artificial neural networks exhibit hallucination-like behavior. His purpose was not to coin the term, but to study the process itself, highlighting how such phenomena can serve as a powerful source of creativity and ultimately form the foundation of his and later generative AI. Periksson28 (talk) 19:18, 17 August 2025 (UTC)

The earliest reference I'm aware of to a computer halliciating is in the 1983 movie: Wargames by Professor Falken in the missle command center. The WOPR computer was depicted as having the ability to learn which is a basic concept of artificial intelligence. Colonial Computer 02:10, 8 August 2023 (UTC)

Move "Terminologies" section earlier?

It seems to me that the "Terminologies" section is foundational and definitional, yet it currently is buried toward the end of the article. I propose moving it earlier, such as before or after where the "Analysis" section currently is. Showeropera (talk) 16:52, 13 July 2023 (UTC)

Wiki Education assignment: Intro to Technical Writing

This article was the subject of a Wiki Education Foundation-supported course assignment, between 3 October 2023 and 1 November 2023. Further details are available on the course page. Student editor(s): Wdan14 (article contribs).

— Assignment last updated by Jazaam02 (talk) 19:28, 8 December 2023 (UTC)

Glenfinnan

Why is the Glenfinnan bridge a particularly "notable" example of this phenomenon, as the lead claims? Furius (talk) 02:17, 23 March 2024 (UTC)

Agreed! Perky28 (talk) 21:52, 21 January 2025 (UTC)

Stephen Thaler in the Origin section

Hey Perky28! You inserted this into the Origin section, and currently edit-warring after two reverts by me and another user:

In 1995, Stephen Thaler introduced the concept of "virtual input phenomena" in the context of neural networks and artificial intelligence.[11][12][13][14][15] This idea is closely tied to his work on the Creativity Machine.[16] Virtual input phenomena refer to the spontaneous generation of new ideas or concepts within a neural network, akin to hallucinations, without explicit external inputs. Thaler's key work on this topic is encapsulated in his U.S. patent "Device for the Autonomous Generation of Useful Information" (Patent No. US 5,659,666), granted in 1997. This patent describes a neural network system that can autonomously generate new information by simulating virtual inputs. The system effectively "imagines" new data, due to a variety of transient and permanent network disturbances, leading to innovative and creative outputs. This concept is crucial in understanding how neural networks can be designed to exhibit creative behaviors, producing results that go beyond their initial training data and mimic aspects of human creativity and cognitive processes.

The first sentence might be relevant, but is sourced by 5 articles by Stephen Thaler, see WP:PRIMARY. Wikipedia requires secondary sources, so to show that Thaler invented Hallucinations in AI you need to find someone else writing about that in reliable sources.

Everything else is completely WP:UNDUE or even WP:FRINGE, and looks like self-promotion. Thaler looks like a patent troll (see ), and the article on his DABUS is also tagged with extensive use of primary sources. Quoting Thaler's website : Thaler is therefore both the founder and architect of confabulation theory and the patent holder for all neural systems that contemplate, invent, and discover via such confabulations, even though all his papers amassed just a bit over 1000 citations in ~30 years according to Google Scholar. The did used the word "hallucination", but as far as I see, he has almost zero influence on modern neural nets development besides his patent wars.

I'm reverting your addition, please bring your arguments here if you must. Artem.G (talk) 12:09, 22 January 2025 (UTC)

Late reply, but the remaining section is still a bit dubious. The actual term the article is about "AI hallucination" never appears in any of the papers, though "hallucination" does. A bunch of papers didn't mention Thaler at all. Cortador (talk) 21:07, 6 February 2025 (UTC)

Spin doctors' anthropomorphic name

If the system was sentient AND self-aware, "hallucinating" might be an appropriate term.

But the system is neither sentient nor self-aware, and the human point of view is much more relevant. If we anthropomorphized more conservatively (without pretending it's a self-aware being that needs psychiatric care), a more appropriate term would be "lying". ("Confabulating" is more psychologically accurate, but only from the system's anthropomorphic point of view which doesn't exist. If you hired a human assistant to summarize the facts about a topic and this is what they brought back, they'd be either lying or incompetent, not confabulating.)

If we didn't anthropomorphize (which I'd prefer), it would be called "generating deceptive garbage", or just "failing".

A name given by people whose livelihood depends on the market success of such systems should not be treated as descriptive or truthful, even when it's the word everyone is using and therefore the right name for the article. OF COURSE they'd prefer to deflect their shareholders' attention away from "garbage" and "failure" - but that's what this really is. TooManyFingers (talk) 14:10, 12 July 2025 (UTC)

It is the term of the art, akin to, say, most cases of COVID-19 being "mild", as in not requiring hospitalization. WP:IDONTLIKEIT & WP:SOAPBOX pertain. kencf0618 (talk) 00:44, 28 July 2025 (UTC)

Does this event belong on this page?

https://www.theguardian.com/technology/2025/aug/12/us-man-bromism-salt-diet-chatgpt-openai-health-information

https://thehill.com/policy/technology/5446903-sodium-bromide-poisoning-chatgpt-advice/ Cahlin29 (talk) 07:42, 13 August 2025 (UTC)

A Commons file used on this page or its Wikidata item has been nominated for speedy deletion

The following Wikimedia Commons file used on this page or its Wikidata item has been nominated for speedy deletion:

• AI Overviews result for "Joaquin Correa brother", 10 August 2025.jpg

You can see the reason for deletion at the file description page linked above. —Community Tech bot (talk) 12:37, 18 August 2025 (UTC)

Wiki Education assignment: Advanced Topics in Digital Culture

This article was the subject of a Wiki Education Foundation-supported course assignment, between 5 September 2025 and 11 December 2025. Further details are available on the course page. Student editor(s): Iamiasc, SwarmOf1Locust (article contribs).

— Assignment last updated by StarryGlow (talk) 19:50, 15 October 2025 (UTC)

Is this really "low" importance?

AI "hallucinations" are a serious problem for everyday people who tend to believe the first thing that LLMs spit out. I've repeatedly experienced errors from every AI from ChatGPT and more esp. Grok. I've even got into verbal altercations with both of them.

Microsoft's ChatGPT extensions also give it a heavy left-wing bias (and I'm left wing myself so I can see when it's really lost the plot).

Some AIs apologise when they are called out, MS's Chat system will often deny ever making the mistake. I challenged Grok on something it told me and it denied that it ever said it despite being an accurate copy/paste of what it said.

In either case, these AIs refuse (or are unable) to update their databases with the correct information. Seems to me (and I'm obviously not a regular user, I find Wikipedia's is often a bit of a "Boy's Club" (and I use that term in a gender neutral fashion) that this should be given far more visibility since we're heading down a rabbit hole that could seriously damage our collective knowledge esp. as these answers are posted back to the net and are subsequently re-absorbed into the LLMs causing the issue to spiral.

It's particularly disturbing when an AI simply throws its toys out of the pram and cuts you off for being called out. 2A04:4A43:919F:F5D0:2B81:245A:8D00:AAEA (talk) 11:20, 2 November 2025 (UTC)

You can change the importance parameter from "low" to "mid". Orientls (talk) 16:03, 2 November 2025 (UTC)

I adjusted the assessment to "mid" for "Artificial intelligence". Alenoach (talk) 00:00, 8 November 2025 (UTC)

@Alenoach and Orientls: Should this article not be assessed as "high-importance?" It describes one of the most fundamental limitations of generative AI. Jarble (talk) 17:13, 8 November 2025 (UTC)

Seems warranted, I would support setting it to high for "Artificial intelligence". Alenoach (talk) 17:34, 8 November 2025 (UTC)

Would it be beneficial to expand on the "Hallucination in Education" section?

Hi! I am a student working on this article for a project and added the section "Hallucination in Education" section a little while ago. I was wondering if anyone thinks there's a viable way to expand on that section. Maybe adding more examples of how it affects the industry, or should I leave it as-is? Thanks! SwarmOf1Locust (talk) 16:52, 26 November 2025 (UTC)

Hi SwarmOf1Locust, thanks for your work. The article is already quite long and there is a risk of repeating things that are already said elsewhere (this article has a tendency to become redundant). You're free to make bold edits if there is something you would like to add, but it may be easier to improve the article by condensing, explaining better or reorganizing the existing content, notably in the "Examples" subsection (which seems too long and detailed) or in "Mitigation methods". Alenoach (talk) 04:08, 27 November 2025 (UTC)

Controversies by country

• In one European country, 2025-December-18, an ai app and/or ai system was used by at least two psychologists to acquire diagnosis of patient; not done according to current laws. Source, https://www.vg.no/nyheter/i/M7gKrm/ai-verktoey-hallusinerte-i-terapirommet , Retrieved now. ~2025-37326-10 (talk) 12:58, 18 December 2025 (UTC)

Removal of content sourced from Hiroko Konishi

I removed the content based on Hiroko Konishi's papers. These are pre-prints that look largely AI-written and are far from reaching the bar for peer-review in any decent journal. Just look at the "References" section, which only cites four of the author's own preprints. There may also be a COI from the contributor to this content, Wecklost23. Alenoach (talk) 02:02, 31 January 2026 (UTC)

I went across the user's contribution history and it's basically the duckiest SPA imaginable. Thank you for posting this and nipping the contrib in the bud. You've probably saved us tens of sapiens-minutes of clean up work!

Update: I was expecting some high-difficulty technical wall-of-text when I opened those paper, and got appalled seeing the EOF at page 8.

The author might need to buy Gemini Pro first before inventing AGI.

iris 11:09p, edited 0:59a the next day. 海盐沙冰 _{/ aka irisChronomia / Talk} 15:09, 31 January 2026 (UTC)

This LLM-generated text has been collapsed and should be excluded from assessments of consensus. 海盐沙冰 _{/ aka irisChronomia / Talk} 15:03, 31 January 2026 (UTC)

The following discussion has been closed. Please do not modify it.

I would like to respond to this removal rationale.

First, the fact that a work is a preprint does not disqualify it from use under Wikipedia policy. Wikipedia does not require peer review as a prerequisite for inclusion. Primary sources may be used for defining concepts and identifying originators, which is precisely how these papers were cited.

Second, characterizing material as “AI-written” based on appearance is a subjective judgment and not a policy-recognized criterion for content removal. No specific passages have been identified, and no verifiable methodology has been presented to support this claim.

Third, the presence of multiple works by the same author in the references section reflects the reality that this concept originates from that author’s research. That alone does not constitute a conflict of interest violation. COI concerns require evidence of promotional intent or distortion, neither of which has been demonstrated here.

If there are specific policy-based concerns regarding neutrality, undue weight, or sourcing balance, those should be addressed explicitly and through targeted revision. Blanket removal based on speculative assessments of authorship quality or hypothetical COI is not consistent with Wikipedia’s content policies. Wecklost23 (talk) 14:33, 31 January 2026 (UTC)

Go read WP:OR, WP:SPA and WP:SPSPREPRINT. 海盐沙冰 _{/ aka irisChronomia / Talk} 14:57, 31 January 2026 (UTC)

In court filings

Hi @Lindenfall, I think the subsection "In legal filings" is well-written but the accumulation of examples of people punished for hallucinated citations can feel a bit repetitive. Maybe we should avoid adding more examples; and some of it may benefit from being moved to more specific articles if there are such. What do you think? Alenoach (talk) 21:30, 21 May 2026 (UTC)

Hi, @Alenoach, and thank you. I couldn't agree more. It's enough. I'd added the Fortune mag sidebar quote as a summation, and placed it up top, hoping it would discourage excess cases being added. The database name is also there. I think just an annual update would do. Alternatively, I considered adding a section to Law practice management software regarding, but not quite a fit, and a little obscure. I added something here as well: Artificial intelligence#Other industry-specific tasks, but that's also obscure. Might warrant an article on Artificial intelligence in law firms. Please jump in to cull or add, if you're inclined. Lindenfall (talk) 21:49, 21 May 2026 (UTC)

Not a big problem, I mostly wanted to let you know, so that you don't spend effort adding more examples. Alenoach (talk) 22:42, 21 May 2026 (UTC)

Over use of preprints

As an encyclopedia, functionally, Wikipedia mostly just summarizes WP:SECONDARY sources. The AI industry's apathy to academic publishing, peer review, journals, editorial oversight etc. is not an excuse to add poor sources to this or any other article. WP:ARXIV and other pre-prints are not generally reliable sources. If a perspective is important, please cite a reliable, independent source for that perspective. If you cannot find such a source, that perspective isn't presumed to be important after all. Also, be aware that people often spam their own works to Wikipedia for clout, and academics and researchers are no exception. Grayfell (talk) 05:56, 31 May 2026 (UTC)

I'm seeing preprints in many technical AI-related articles. WP:SPS says "Self-published sources may be considered reliable if published by an established subject-matter expert, whose work in the relevant field has previously been published by reliable, independent publications. Be careful when using such sources: if the information in question is suitable for inclusion, someone else will likely have published it in independent, reliable sources." That's vague, and I agree that using clearly published sources would be preferable. However, this could be an opportunity to develop new guidelines about establishing expertise.

Requiring in-prose attribution for acceptable preprints could be a component. I would rather not be deletionist about verifiable technical information just because likely-expert authors didn't prioritize publishing, but I know others may disagree. There are so many journals now, I can't even tell if a journal is reliable by just seeing its title. I may need to dig through the website to find a "peer-review" claim which I can't immediately verify. I don't necessarily want to add more incentive for scientists to "get published" by arbitrary journals rather than evaluating reliability in a more up-to-date way. I'm fine with excluding preprints from GNG for technical articles. WeyerStudentOfAgrippa (talk) 16:40, 31 May 2026 (UTC)

"Deletionist" is a loaded term. If reliable sources talk about this information, we can figure out how to include it so that disinterested readers will be able to understand it. Unfortunately, that's not what's happening for AI articles. Many of these preprints are made by academics who at least have some degree of accountability with a research institution, but a lot of them are now part of AI companies. VC doesn't seem to care about publication, so these preprints functionally exist for promotional purposes. Just as we shouldn't cite product listings or press releases for the latest and greatest products, we shouldn't cite preprints to decide which new technique is important enough to mention in an overview of a massive, complicated topic.

Determining which journals are reputable and which are not can be tricky, but preprints have not yet been published in any journal, so that's a separate issue. As an encyclopedia, articles aren't intended to advance the cutting edge of research. We're trying to explain a complicated topic to a general audience. If nothing else, if a preprint is the only source we can find for something, that should be a sign that we've lost the plot.

I've seen it implied that AI is so new and develops so rapidly that publishing is obsolete, or is an unrealistic burden. From this perspective, AI should be allowed to bypass basic standards of fact-checking, due weight, accuracy, etc. This is more a product of industry hype and mass AI psychosis than it is about the reality of AI research. Regardless, Wikipedia isn't an arm of the AI industry. If the AI industry wants to be treated like an industry, we should avoid flimsy commercially-motivated sources. If it wants to be treated like an academic discipline, it should be held to that standard, instead. Either way, these articles have a lot of crappy sources. Grayfell (talk) 22:22, 31 May 2026 (UTC)

Grayfell, I don't think the source flagged in this edit is a preprint, it was published in ACM Computing Surveys. It also has over 8000 citations, so notability shouldn't be a problem. Alenoach (talk) 03:53, 1 June 2026 (UTC)

You're right, my apologies. For convenience, here's the paragraph:

Ji et al. divide common mitigation methods into two categories: data-related methods and modeling and inference methods. Data-related methods include building a faithful dataset, cleaning data automatically, and information augmentation by augmenting the inputs with external information. Model and inference methods include changes in the architecture (either modifying the encoder, attention, or the decoder in various ways); changes in the training process, such as using reinforcement learning; and post-processing methods that can correct hallucinations in the output. (doi:10.1145/3571730)

The paper divides these into categories as a way to explain them in detail. But this paragraph doesn't do that, it just says that the categories exist with a tautological definition: Data-related methods are methods that work with the data, while model and inference methods work with models and inference. I don't think this is very informative, and as I've made clear, I think these article are already bloated.

I admit I haven't read the source closely (obviously) but it does look like it might be useful for more basic information. As one example, section 2.1 explains the difference between intrinsic and extrinsic hallucinations. IMO the article would benefit from explaining these kinds of fundamentals more clearly, and this source might be more useful for that. Grayfell (talk) 04:39, 1 June 2026 (UTC)

The first sentence of the paragraph looks unnecessary and I wouldn't mind if it is removed. Alenoach (talk) 20:36, 1 June 2026 (UTC)

Oh, also, neither citation count nor the notability of the paper itself make it reliable. Many bad sources have very high cite-counts, and many good sources are obscure. Grayfell (talk) 04:40, 1 June 2026 (UTC)

For such a rapidly evolving new science this article relies too much on old content

This article came up here: https://en.wikipedia.org/w/index.php?title=Wikipedia:Village_pump_(policy)&oldid=1365342472#Telling_modern_paid_GPT_5.6_to_get_me_AI_hallucination_sources

Where I found ten sources apparently via GPT not in use here. Some are older. I ran the article through to see how current the sourcing on hallucination current realities are, and was surprised to see so much of the article sourced to older things still -- and likely outdated?

Sources by publication year

Year	Number of sources	Number of citations	Approximate percentage
1999	1	1	0.7%
2007	1	1	0.7%
2008	1	1	0.7%
2014	1	1	0.7%
2015	2	2	1.5%
2016	1	1	0.7%
2018	2	3	2.2%
2019	3	4	2.9%
2020	2	2	1.5%
2021	2	2	1.5%
2022	15	15	11.0%
2023	28	35	25.7%
2024	16	26	19.1%
2025	23	25	18.4%
2026	14	15	11.0%
Undated	2	2	1.5%

Those sources may be useful at that link here. And this article needs updating to reflect whatever this actual current latest science is, not predominately how it looked circa 2022-2025. — Very Polite Person (talk/contribs) 20:33, 21 July 2026 (UTC)

Have you found claims in the article that are inaccurate? WeyerStudentOfAgrippa (talk) 00:04, 25 July 2026 (UTC)

there are still numerous claims on this article that are both inaccurate and or compressed the nuance and rely on older generations of technology.. I'm still doing research myself and I want to eventually edit some things, as I just discovered this article. PlutoidRepublic (talk) 19:09, 31 July 2026 (UTC)