Talk:Dead Internet theory

RfC

I don't think we're going to come to an agreement here, so the best option is to hold an RfC. I see that some have been held on this previously, but they didn't consider the sources that did not consider it a conspiracy so I feel it is appropriate to hold a new one.

I propose:

Should this article:

A: Describe the Dead Internet theory as a "conspiracy theory" in Wikivoice, and present it as a conspiracy theory that proposes that since 2016 the internet has consisted primarily of bot activity and automated content manipulated by algorithmic curation, that aims to control the population and reduce genuine human interaction. The non-conspiratorial alternative can be mentioned in the article but will not be the focus. B: Describe the Dead Internet theory as a "theory" in Wikivoice, and present it as a theory that proposes that the internet is becoming predominantly populated by AI-generated content in the wake of generative AI. The conspiratorial origin can be mentioned in the article but will not be the focus.

Do you have alternative suggestions for wording? BilledMammal (talk) 07:15, 13 April 2026 (UTC)

The main issue with "alternative suggestions" for wording is that they are not reflective of the history of sources, or comprehensive to the entire theory. Specifically, if we drop the "conspiracy" we need to drop a large portion of the content. I've been working on a table to aggregate the sources together in my sandbox:

— Preceding unsigned comment added by GeogSage (talk • contribs) 14:21, 13 April 2026 (UTC)

Comment - we don't need to drop "conspiracy", and instead could describe it as a historical progression. Something like:

Dead Internet theory, a term that emerged in the 2010s, originally referred to a conspiracy theory - the idea that the majority of internet activity (content creation? traffic? ) consists of bot activity and automated or algorithmically-manipulated content. In the 2020s, the rapid development of large language model (LLM) AI has lead some writers (researchers?) to use the term to describe a real-world observed trend towards increased bot activity, including algorithmic feeds promoting AI Slop, that could potentially eclipse human-generated content.

But this would necessitate restructuring the article into conspiracy / observed trends / theorized future .Dialectric (talk) 15:40, 13 April 2026 (UTC)

The issue is that a bunch of clickbait articles use it as an attention getter for "lots of bots online." The DIT is still a conspiracy theory according to most sources that are actually discussing the theory itself. In academic literature, there is a letter to the editor in a journal that has been picked up by a few other articles that brush against the topic while discussing other things, but that is only one thread in the literature. These ignore the entire 2nd half of the theory. There is probably a case that the DIT is becoming a colloquial slang for "a lot of bots online," but there isn't a source I've seen on the evolution of the term in the English language. A Nation-state is a very precise term that is often mistakenly used to refer to any sovereign state, I believe DIT is being used in a similar manner, and to get attention on clickbait articles that are regurgitating AI slop about AI slop. We would essentially be restructuring the article based on the sources that are not discussing the topic in enough detail to even get into the original meaning and just using it to bring "attention to a legitimate problem." Most sources don't say it isn't a conspiracy theory, they do some spooky, clickbait, dribble approach of saying something along the lines of "started out as a conspiratorial joke, but it is edging ever closer to reality." Very few go all the way to endorsing it, and the ones that do don't mention the "full" theory that is discussed in other sources, just the part about there being a lot of bots online. GeogSage ^{(⚔Chat?⚔)} 22:40, 13 April 2026 (UTC)

Very few go all the way to endorsing it, and the ones that do don't mention the "full" theory that is discussed in other sources, just the part about there being a lot of bots online.

The theory has evolved. The "full" theory is now just that there are lots of bots online - the article should reflect that. It can mention the claims of the old theory, but it shouldn't present it as if those claims are what people are referring to when they discuss the Dead Internet theory. BilledMammal (talk) 05:06, 17 April 2026 (UTC)

I've opened the RfC below. I agree with mentioning the historical progression, but not in the first sentence of the lede, as that would be undue focus. BilledMammal (talk) 05:04, 17 April 2026 (UTC)

New source published today ;

The dead internet theory proposes that much of online activity is no longer human — that bots, automated content and algorithmic feedback loops have discreetly taken over genuine interaction. No mention of conspiracy theory.

Some more news sources

states Over the following decades, as the “dead internet theory” took off (positing that our current internet is composed primarily of content generated by bots) and artificial intelligence tools helped individuals or bots proliferate AI slop, the term zombie internet began to be repurposed

states Dead internet theory is the hypothesis that the web either already is or soon will be entirely dominated by artificial intelligence Katzrockso (talk) 00:48, 20 April 2026 (UTC)

Survey (conspiracy)

• B. I conducted a review of the sources on the first six pages of this Google scholar search result, which considers sources from 2025 and 2026, and found that a majority of sources do not consider it a conspiracy theory, explicitly considering it valid. Almost all of the rest considered it to have begun as a conspiracy theory or to have been considered a conspiracy theory, but to now be valid; only two or three of the sources I assessed considered it to still be a conspiracy theory.

As such, our article currently violates WP:DUE and WP:BALASP. The notion that the Dead Internet theory is still a conspiracy theory is a fringe view, but we currently call it one in Wikivoice, and we prominently include conspiratorial aspects of the original theory in the first sentence of the lede, aspects that recent sources give little or no weight to.

This doesn't mean we will call the theory true - while some sources support this, most consider it to just be a theory or projection - and it doesn't mean that we can't mention the conspiratorial origins of the theory, just that we will reflect the current scholarship in the field rather than past scholarship, in line with the requirements of WP:AGE MATTERS. BilledMammal (talk) 05:02, 17 April 2026 (UTC)

I have not gone through all these sources, but have a table linked above to several that we use in the article. You have definitely included several that have already proven to be predatory journals. A blanket review of Google Scholar is not really a good approach. I believe you were a bit generous on some of your assessments. GeogSage ^{(⚔Chat?⚔)} 06:46, 17 April 2026 (UTC)

The only source you previously mentioned as being predatory was "The Dead Internet Theory: A Survey on Artificial Interactions and the Future of Social Media", but when I asked for details as I was unable to find the proof you mentioned, you didn't respond. I'll add that even if it is, it has been cited 20 times and the authors are academics at reputable institutions like the University of Texas, so it would be usable under WP:SPS.

As for the blanket review of Google Scholar, we need a way to review an unbiased selection of sources, as limiting the review to sources currently in the article risks cherry picking. BilledMammal (talk) 12:00, 17 April 2026 (UTC)

The talk pages of this have had many people trying to scrape and cherry pick every source possible to advance the POV that the DIT is not a conspiracy, while ignoring every source that is inconvenient. I've been busy, and a bit stressed out, and honestly repeating myself to everyone pushing the same sources, making the same arguments, is exhausting. Doing so has got me nothing but accusations of tendentious editing or brigading. To answer your question: That source was taken to the reliable source noticeboard here. Summarizing the discussion, it is published by an organization listed as predatory, and based on one AI detection software, was 95% AI-generated. The fact "academics at reputable institutions like the University of Texas" cited it does not give credibility to the article, it undermines the credibility of those authors publications, and the journal that published it, the reviewers who claim to have read it, and the editor who approved it. Great example of why the Argument from authority is a fallacy.

I've been periodically reading sources on the DIT for years now, it is a very common topic for garbage, click-bait, AI-Slop publications that misinform and misrepresent the theory to people by half explaining the concept with sensationalist language. There are so many venues to publish across the world, it is impossible to know them all, and therefore necessary to start from the assumption that everything is a poor source until otherwise demonstrated. Each source evaluated takes a bit of work, which I've been doing and including sources that pass in the article. I was doing this on the table linked above that I assume you used as inspiration for this. You made this table pretty fast, probably because you spent almost no time actually checking if the sources were any good. Going through, one at a time, to do so is not something I have time to do at this moment, so the damage is done. A quick glance though, one other example of a problematic source is The Synthetic Consumer Singularity of a "Dead Internet": Quantifying the Collapse of Attention-Based Business Models Under High-Fidelity Agentic Noise. This is a conference paper published on ResearchGate, not a peer-reviewed journal. That essentially a self published paper, anyone can upload on ResearchGate. You have at least one paper that is published on arxiv, which unless it is published elsewhere, are essentially self-published. Reliable source examples goes into both conference abstracts and arXiv papers. Google Scholar is only good at narrowing the results, it doesn't filter out everything, so it takes work before you accept a source. None of that work was done, and it is presented in a RfC to editors who will take the sources at face value, and requires hours of work for anyone to disagree. Hopefully, sources are actually evaluated and this isn't decided by a vote. GeogSage ^{(⚔Chat?⚔)} 05:12, 19 April 2026 (UTC)

I have to say, I am not a big fan of you casting WP:ASPERSIONS about me and other Wikipedia editors who have looked at the available evidence and disagree with your conclusions. Especially when you start with an easily-checkable lie ("...many people trying to scrape and cherry pick every source possible to advance the POV that the DIT is not a conspiracy, while ignoring every source that is inconvenient.") Name three Wikipedia editors who have argued that "DIT is not a conspiracy". Now name three Wikipedia editors who have argued that "DIT is not a conspiracy theory", which is what I suspect you meant but were distracted by your insulting pretty much everyone else in the discussion.

Wikipedia doesn't say whether something is or is not a conspiracy. Wikipedia doesn't say whether a conspiracy theory is true or false. We simple report what reliable sources say. Some sources (mostly older) say that it is a conspiracy theory. Other sources (mostly recent) say that some percentage of what is on the Internet is indeed "Dead" (not created by humans). I have never seen anyone outside of 4chan or Infowars claim that the percentage is 0% or 100%. --Guy Macon (talk) 10:04, 19 April 2026 (UTC)

Wikipedia:NOTDEMOCRACY. I'm not a big fan of needing to point out bad sources. Go through the articles talk page, and see how many editors over the past several years have made posts on this where I've had to point out the problems with their sources. It is exhausting, and the argument that a source that has already been shown to be predatory and likely 95% AI generated is somehow a RS because researchers at Universities can't be bothered to check their citations was a bit much for me. I don't consider scraping Google Scholar for sources published since 2025 and including ones only on Arxiv, conference papers only published on ResearchGate, and articles that have already proven to be predatory "looking at the available evidence." I can point to several reliable sources from 2024 and 2025 that refer to it as a conspiracy theory, they are cited in the article already. Wikipedia relies on reliable sources, and my problem is that the above "review" is comprised of a hastily generated list of unvetted stuff that has multiple demonstrably unreliable sources in it. Most things that discuss the DIT do not even claim to deeply discuss the it, and do not claim to offer a comprehensive overview or definition. Including a list of sources that includes several not suitable for use in the main space as evidence in a RfC is a problem. The energy to check these sources is outsourced to others, and even when that has is done, it is ignored unless the evidence is repeated on demand every time the source is introduced.

Sources in the table are also not all accurately represented. For example, Generative AI is Breathing New Life Into the Dead Internet Theory discusses "A “Leaner” DIT," which is probably the first time a split is being clearly identified in the literature. The table states the article asserts it was "previously" a conspiracy theory, when it is not quite that straight forward. Within the text it states:

We must admit that some of the core principles of the DIT are convergent with our technical and historical experience. Unfortunately, the conspiracy theorists augment these very plausible observations with their own mix of biases and agendas that lead to implausibility and absurdity, which in turn leads to rejection. But it is a mistake of the first order to dismiss the core criticisms unequivocally.

Essentially, the article is pointing out the same thing James Felton stated in the 2024 IFLScience article, or that Carlos Diaz Ruiz discussed in Market-Oriented Disinformation Research: the DIT is rooted in observable phenomena which is why it is successful as a conspiracy theory. GeogSage ^{(⚔Chat?⚔)} 05:55, 20 April 2026 (UTC)

Thank you, I think I missed that for the same reason you did, though I believe the source is still usable for the reasons I outline above. However, let's assume it isn't - and let's assume that half of the sources I presented that don't support labelling it a conspiracy are also unusable, and that none of the sources I presented that do are.

We still have five times as many sources that do not call it a conspiracy theory as do; at this point, the evidence against is just too strong for us to continue declaring it a conspiracy theory in Wikivoice. BilledMammal (talk) 02:55, 22 April 2026 (UTC)

The source is in a predatory journal (WP:QUESTIONABLE), failed at the reliable source noticeboard, and is likely 95% AI slop (WP:RSML). Starting the 3rd RfC on this exact topic in less then a year (The last RfC was closed by @Bon courage in February with the statement "It is unhelpful to re-litigate matters so recently decided.") with a table filled with sources that do not meet the standards we need for inclusion, and then leaving the table and problematic sources up during the discussion after they've been pointed out, is a real mess. I'm not sure how many are actually usable, but would be surprised if half are from what I've seen. I don't agree with your "Conspiracy Theory?" columns conclusions, I think you were much to liberal in your interpretation of sources thinking it either is not or was previously a conspiracy theory. You also limited to 2025, which ignores the source that established the DIT in the literature and many other sources that go into detail on the DIT, not just mention it in passing in a discussion of AI bots. The table I had been working on is not comprehensive of the ones that are used in the article, I was trying to take my time and check the reliability of them, and there are sources that call it a conspiracy theory not included in the article because they are redundant to ones we cite.

What we have is multiple sources that declare it a conspiracy theory (including a very clear glossary definition, as well as the oldest reliable sources on the topic), multiple that only focus on the observations made in the first part of the theory, and a source that describes the second group as a "leaner" version of the extracted from the DIT. The third source is one I've been expecting to see for a while, and it neatly cleans up and contextualizes the fork issue discussed by the closer of the first RfC closed on December 29th 2025. This source does not state the lean version replaces the original DIT. In it's current form, the article represents the full picture of the literature as far as I can tell. A metaphor that comes to mind is the Coca-Cola page where Diet Coke and New Coke are mentioned in the lede, but replacing Coca-Cola with these as the main focus of the article would be odd even if there was suddenly a few more recent sources discussing these products. GeogSage ^{(⚔Chat?⚔)} 05:32, 22 April 2026 (UTC)

AI detectors are notoriously unreliable, especially for scholarly works. Reading it, I personally don't see indications it was AI written - and I did exclude a few sources because they appeared AI written.

I limited my survey to post-2024 because WP:AGEMATTERS. This is a fast moving field, and sources that were useful three years ago are no longer suitable; recent sources, with very few exceptions, decline to label it as a conspiracy theory, and prefer to treat it as a valid concept. BilledMammal (talk) 05:44, 22 April 2026 (UTC)

The source is not acceptable for use on Wikipedia for two reasons, it is a predatory journal and was flagged as likely AI generated. There are other sources that are problematic as well in the table, for example Who Are You Talking To? The Discernment of AI and Human Content on Social Media failed the noticeboard here. There are multiple masters thesis, at least one dead link, and several others that are self published on sites like ResearchGate. Leaving the table while these sources remain and using it as a point in discussion is problematic.

We have a 2026 source that explains the discrepancy between these sources. Importantly, it directly addresses the opinion piece that has been cited by several articles that don't focus on the comprehensive theory. As stated, I disagree with your interpretation of many of the sources as well. GeogSage ^{(⚔Chat?⚔)} 06:04, 22 April 2026 (UTC)

Multiple masters thesis? I reviewed the list again and found I did miss one, which I have now removed, but I couldn't find any more. I also fixed the two dead links and removed a few other sources, including every source you objected to - while I don't agree with your assessment in every case, the weight of sources against using "conspiracy theory" in Wikivoice is too great for it to be worthwhile arguing over individual sources.

If there are further sources you disagree with, please say - even with these sources removed, there are still 7 times as many sources that don't consider it a conspiracy theory as do. BilledMammal (talk) 06:44, 22 April 2026 (UTC)

Thought I counted two thesis, if not my mistake. That said, you need to assume the source is unreliable, and prove it is acceptable before using it, rather then waiting for it to be challenged. It is outsourcing the labor to others to check your work. This isn't just a problem for this, it is important to learn to vet sources. For example:

• THE DEAD LAW THEORY: THE PERILS OF SIMULATED INTERPRETATION is a pre-print hosted on SSRN. Until this is published somewhere, it is essentially self published. WP:PREPRINT
• The internet is dead – Updating metaphors of the web for the 2020s is a conference abstract. I'm not quite sure what the bar to enter that conference was, but generally these are lower quality sources. I've presented at a lot of conferences, and my abstracts were never checked by anyone but me.
• AI-generated Images of Ukrainian Soldiers as a Tool for Media Manipulation in the Context of the Russo-Ukrainian War is published in Journal of Cultural Analysis and Social Change. This journal was discontinued from SCOPUS in 2024 according to SCImago Journal Rank. My understanding from a quick search is that this was because of predatory practices.
• Analysing Multidisciplinary Approaches to Fight Large-Scale Digital Influence Operations is a self published Arxiv link to a conference paper that won't even be presented till September 15, 2026 (assuming the uploader is accurate in saying it was accepted). WP:PREPRINT
• AmongOthers: A Design Speculation for Rethinking AI in Online Social Communities is a conference paper. I'm not sure about the review process for this conference, they vary, but this seems to be a more reputable conference. This conference happened a few days ago I believe.
• Somebody’s Poisoned the Water Hole! Advancing Social Media Data Collection Principles to Account for the Effects of AI Slop on Public Relations Scholarship is a pre-print hosted on SSRN. Until this is published, it is essentially self published. WP:PREPRINT

Of these, the strongest case could probably be made for AmongOthers: A Design Speculation for Rethinking AI in Online Social Communities as I think that is a reputable conference based on the Wikipedia page for it, but I generally avoid conference proceedings unless I can help it. I would not let my students cite these in their thesis. When it comes to preprints, I avoid them like the plague because it can change dramatically when finally published, and you don't know where it will end up. Citing something in pre-print form in a document can lead to embarrassment if the final version no longer confirms what you're using the citation to back. GeogSage ^{(⚔Chat?⚔)} 23:58, 22 April 2026 (UTC)

I've tried to go through and review most of the sources that were included in this to see if they can be included in the article text. It represented a significant amount of work as there is a lot of questionable literature on this topic. While I'm not finished, I'm going to include a collapsed version of the table I had been working on covering relevant sources and quotes in the RfC to make sure the closer and others have it included. I believe that opening a third RfC in less then six months with dubious sources, and not removing all of the sources that have been pointed out as dubious after they were included, is really not good practice. One issue that should be pointed out is that many sources only briefly namecheck the DIT and don't go into much detail, much less attempt to claim a comprehensive definition, making it hard to infer what the author actually thinks about it. This is a possible WP:REFBOMB issue. GeogSage ^{(⚔Chat?⚔)} 23:00, 26 April 2026 (UTC)

• A The 'Dead Internet theory' is not the same thing as 'there are lots of bots online'. TarnishedPath^talk 07:28, 17 April 2026 (UTC)

  It didn't use to be, but my source review shows that the definition has changed over time and become something close to 'there are lots of bots online' BilledMammal (talk) 12:00, 17 April 2026 (UTC)

• C: Begin with the conspiracy theory origin and then describe how it has recently begun to shift toward an exploration of the proliferation of slop-bots. The source review is somewhat compelling but is limited by WP:RECENTISM issues. I think it's pretty clear that right now "Dead Internet Theory" is taking on a new meaning both online and in the academy but it would be doing a disservice to readership to downplay its conspiratorial origins. We should not be taking a POV on the meaning of this topic and should, instead, describe its evolution. Simonm223 (talk) 13:13, 17 April 2026 (UTC)

  coming from NPOVN. C) describe both the original conspiracy theory and the current descriptor in some cases as proliferation of gen-AI

  seems like a fair way to handle it. actually, could possible be due for a split article eventually if the two are completely different User:Bluethricecreamman (Talk·Contribs) 14:21, 17 April 2026 (UTC)

• NONE OF THE ABOVE / BAD RFC: Describing it as a conspiracy theory in Wikivoice ignores the many sources that use the phrase to describe an unproven (but not conspiracy) theory that the majority of the content on the Internet is generated by non-humans. Merely mentioning the conspiracy theory ignores the many sources that are indeed pushing a conspiracy theory that everything on the Internet is generated by non-humans and that this is purposely done to control people's minds. Both exist. No reliable source even tries to estimate which meaning is more common. So simply present both meanings without picking one and say that the phrase has two common meanings. --Guy Macon (talk) 15:36, 17 April 2026 (UTC)

  Would option C (not listed in the RfC, but described by Simonm223) suffice for your position? Katzrockso (talk) 20:04, 17 April 2026 (UTC)

  Yes, but only if "Begin with the conspiracy theory origin" avoids saying anything about what it "is" or "was" in wikipedia's voice. We should not define phrases that never became widely used in Wikivoice. That's for established facts. Someone trying to assign a meaning that never became widely accepted should be attributed: say who said it and say when and where they said it. Don't present it as the only definition or even as the only definition at a particular time. --Guy Macon (talk) 23:20, 26 April 2026 (UTC)

  For what it's worth, I agree with you and am trying to work in that direction. Certainly I generally prefer attribution. Simonm223 (talk) 23:26, 26 April 2026 (UTC)

  Also my early suggestions about chronology were not yet refined. If you look at the first draft I suggested it put up the concept first and referred to the AI related rhetoric as just that. Simonm223 (talk) 23:28, 26 April 2026 (UTC)

  Wikipedia:Call a spade a spade. The full theory has reliable sources ranging from 2021 to this month that call it a conspiracy theory. We have several more recent sources that refer only to what is described as the lean version, or that only mention the DIT in passing without taking a stance on the full theory. There are few if any reliable sources that declare that the full theory is not a conspiracy, the ones that refer to it as either no longer a conspiracy or as increasingly plausible are generally referring to only the core evidence, not the actual thesis of the theory. We now have a source that distinguishes between the full and lean versions of the theory and the article now currently reflects that source in the lede and body of the article. I believe it resolves the discrepancy in the literature pretty cleanly. To discuss the entire DIT and reflect the sources, we should not start by focusing on the more recently emerged subset, or allow the existence of the new lean version that emerged from the original DIT to obfuscate the fact we have multiple reliable sources (including at least one published in a peer reviewed journal in April of this year) in various types of publication (academic journals, new articles, magazine articles, policy papers, books) that use the words "conspiracy theory" to describe it. Diluting the description of the original full theory because a new subset exists is not reflective of the literature in my opinion. GeogSage ^{(⚔Chat?⚔)} 00:07, 27 April 2026 (UTC)

• Option C, for now. Once sourcing continues to develop, maybe in a year or so, we might be able to change the wording to not start with the conspiracy theory aspect, as it will be largely historical and would be UNDUE to lead with. Katzrockso (talk) 20:04, 17 April 2026 (UTC)
  • I have no opinion as to which should comes first. I just want to avoid any hint that there is only one accepted definition for the phrase. Both meanings should be mentioned prominently in the lead. --Guy Macon (talk) 02:27, 18 April 2026 (UTC)

    For what it's worth, my C proposal was for basically this reason. Simonm223 (talk) 19:48, 18 April 2026 (UTC)

• Option C, as described by Simonm223, in line with my comment/suggestion in an earlier section on this talk page.Dialectric (talk) 22:17, 17 April 2026 (UTC)
• Anything other than A. There is a high bar to call something a conspiracy theory in wikivoice. The sources need to be ironclad. Thebiguglyalien (talk) 18:49, 18 April 2026 (UTC)
• Option C. Describe both the "original conspiracy theory" AND the post-AI boom colloquial usage describing AI slop bots, per reliable sources. This is what the article lead already currently does! PK-WIKI (talk) 21:26, 18 April 2026 (UTC)

  @PK-WIKI the lead sentence starts off with "The dead internet theory is a conspiracy theory". Option C would change this to not lead with the false and univocal assertion that the DIT is a conspiracy theory, but rather that some sources described it that way. Katzrockso (talk) 02:15, 19 April 2026 (UTC)

  Option C above says "Begin with the conspiracy theory origin and then describe how it has recently begun to shift toward ... slop-bots." which is exactly what the current lead does. Nothing about the Option C I voted for says we would not lead with the original conspiracy theory. Nothing about Option C mentions changing the verbiage that it is a conspiracy to "some sources describe it that way". PK-WIKI (talk) 05:07, 19 April 2026 (UTC)

  @Simonm223 can you clarify what Option C might say. Would Option C lead to the article still retaining its opening sentence? The dead Internet theory is a conspiracy theory that asserts that, since around 2016, the Internet has consisted primarily of bot activity and automated content manipulated by algorithmic curation I interpreted it as contradicting this, because you replied to Guy Macon's comment that says Both meanings should be mentioned prominently in the lead with For what it's worth, my C proposal was for basically this reason, as well as your initial comment. Katzrockso (talk) 00:35, 20 April 2026 (UTC)

  For what it's worth, I don't think Begin with the conspiracy theory origin and then describe how it has recently begun to shift toward an exploration of the proliferation of slop-bots. means State that the theory is a conspiracy theory in wikivoice in the lead sentence. Katzrockso (talk) 00:36, 20 April 2026 (UTC)

  If I was to take a stab at a first draft I might say Dead Internet Theory is a concept that asserts that, since 2016, the Internet has consisted primarily of bot activity and automated content manipulated by algorithmic curation. Originally constituted as a conspiracy theory, this concept has gained renewed attention as a rhetorical device for describing the impacts of generative AI.

  However I will concede this is really just a first try. Simonm223 (talk) 11:39, 20 April 2026 (UTC)

  That sounds like the general direction that I was interpreting your option C as. Katzrockso (talk) 18:03, 20 April 2026 (UTC)

  The "original creepypasta from 2016" is and will forever be a conspiracy theory, and we must state that in wikivoice.

  That is currently the first sentence / first paragraph of the lead. I would be fine with the "original conspiracy" being the second paragraph, and the "post-AI boom reality" being the third paragraph. But that would require a good first paragraph that melds the two and describes it in a very generic way without getting into the details that will be mentioned in the next paragraphs. I haven't seen one yet. PK-WIKI (talk) 06:58, 21 April 2026 (UTC)

  I agree with the fact that we need to state the original creepypasta or whatever was a conspiracy in wikivoice, but I object to the statement "the dead Internet theory is a conspiracy theory" in wikivoice. Katzrockso (talk) 11:59, 21 April 2026 (UTC)

  I strongly agree with the above. --Guy Macon (talk) 15:31, 21 April 2026 (UTC)

  While the original conspiracy theory said since 2016, most of the sources that treat it as valid say "since Gen AI" or similar. How about:

  Dead Internet theory begun as a conspiracy theory claiming that since 2016 the internet had consisted primarily of bot activity for the purpose of controlling the human population. Since the advent of Gen AI, it is has come to refer to the replacement of genuine human interaction with synthetic content online.

  BilledMammal (talk) 00:57, 22 April 2026 (UTC)

  The first sentence should introduce the subject, not delve into the origins of the topic, per MOS:FIRST. Katzrockso (talk) 03:57, 23 April 2026 (UTC)

  Strongly agree. We need to be careful not to assume that what a bunch of Wikipedia editors are debating is thus important to the reader. The above would be a good addition to the origins and spread section. --Guy Macon (talk) 14:00, 23 April 2026 (UTC)

• Option C - Coming from NPOVN & I agree with Simonm223's proposal as a good way to handle the apparent dichotomy between the two uses of the term.
  Anecdotally, I must admit that I was previously unaware that the "Dead Internet theory" was at all related to any conspiracy theory, having personally only heard it used to refer to the excessive number of bots clogging-up social media platforms. Guess you learn something new every day. - Butterscotch Beluga (talk) 23:12, 19 April 2026 (UTC)
• B or C. Always thought it was a bit cute we described it as a conspiracy theory, mainly on the initial basis of, apparently, some Atlantic article said so. But with the proliferation of generative AI it's become tone-deaf, especially since said theory requires no "coordinated" conspiracy. VintageVernacular (talk) 05:21, 20 April 2026 (UTC)
• Option B or C the majority of current reporting/sources does not characterize this as a conspiracy anymore. --Ita140188 (talk) 13:48, 20 April 2026 (UTC)
• Based on the literature, option A is really the only viable choice. A new source finally exists in a peer reviewed journal that differentiate between the conspiracy theory, and the "leaner" version of the theory without the conspiracy elements. The leaner version of the theory is contextualized in relationship to the full conspiracy theory as one which has spun off of the original one in the literature. Result is that the sources using it to mean "a lot of bots online" can be lumped into the lean version. There is no longer really a discrepancy as it has been cleanly described in the literature. I've already incorporated the source a bit, and restructured some of the article to reflect this.

GeogSage ^{(⚔Chat?⚔)} 14:01, 20 April 2026 (UTC)

How does that make Option A the only viable choice? Option A: [...] The non-conspiratorial alternative can be mentioned in the article but will not be the focus. With the "original conspiracy" and the "AI slop reality" now co-existing in the literature to your liking, something like Option C above would seem to better match the situation. PK-WIKI (talk) 06:39, 21 April 2026 (UTC)

The differences between option A and C are pretty trivial, however the new source does not say that the DIT has "has recently begun to shift toward an exploration of the proliferation of slop-bots." The source distinguishes "between the conspiracy-laden DIT, and a leaner version stripped of paranoia, prejudice, politics and polemic." It goes on to state "So, if we extract from the DIT all of the conspiracy theory-laden baggage, it would appear that there may well be something to be learned by looking at a leaner version of it." This is not a replacement of the original meaning, it is a subset of it that still recognizes that the original exists, stating: "We must admit that some of the core principles of the DIT are convergent with our technical and historical experience. Unfortunately, the conspiracy theorists augment these very plausible observations with their own mix of biases and agendas that lead to implausibility and absurdity, which in turn leads to rejection. But it is a mistake of the first order to dismiss the core criticisms unequivocally." The conspiracy is where the extract comes from, and we do not have a source to the best of my knowledge that is mentioning "colloquial usage," much less stating that the colloquial usage is replacing the full theory. We also have multiple reliable recent sources that discuss the full theory. A bridge just exists to show people are looking at the DIT, and acknowledging the observations made while rejecting the conclusion. GeogSage ^{(⚔Chat?⚔)} 14:26, 21 April 2026 (UTC)

There are many, many sources focusing on what the source above calls the "leaner" version, aka the "no-conspiracy AI boom slop bot proliferation". Those sources are the "colloquial usage" of using the phrase "dead Internet theory" to describe that leaner version. Both it and the "2016 creepypasta conspiracy" version must be described in the article. One evolved from the other and the article should describe the timeline and evolution. PK-WIKI (talk) 19:57, 21 April 2026 (UTC)

We have many, many sources focusing on the full version of the theory as well, and now we have a source that actually discusses the discrepancy. Both the original version and the lean version exist simultaneously, the lean version focuses on the core observations and evidence used in the full. Most sources do not go into detail about definitions or expand upon what the author actually is talking about. It did not evolve from the DIT, it is extracted from the full DIT. We can and do discuss the leaner version, but to do that we must discuss the full version that still currently exists in parallel, and that has sources going back for half a decade at this point. Many of the sources focusing on the leaner version are not ones we can use on Wikipedia (lots of predatory journal articles and self published content), and several others are permissible in some circumstances but not the best (I think there are two masters thesis linked in the table for example, which are "considered reliable only if they can be shown to have had significant scholarly influence"). Why would we start with and emphasize a new subset instead of the existing full DIT that exists within the literature? GeogSage ^{(⚔Chat?⚔)} 04:38, 22 April 2026 (UTC)

• B or other. "Conspiracy theory" has always been a silly term unsuited to serious use by an encyclopedia. It doesn't really convey much information; either something is true or it's false, and being a conspiracy doesn't have much bearing on that. Watergate was a conspiracy but it actually happened; George Washington and the cherry tree wasn't a conspiracy and didn't actually happen. The only thing it really conveys is that the person saying it thinks the thing is stupid. It may really be stupid, but our articles do not call Stalin a tyrant or Nixon a crook, even though those things are true. There are some things that are so widely and universally understood to be obviously false and stupid, and also widely and universally called a "conspiracy theory", that it may be unavoidable for us to do the same, but these are fairly uncommon, and this is not one of those. jp×g🗯️ 21:44, 22 April 2026 (UTC)

  @TurboSuperA+: jp×g🗯️ 23:17, 22 April 2026 (UTC)

  Conspiracy theory is a well known term that is studied in academia and has a well understood meaning. And there are examples of conspiracy theories which end up being true or which have uncertainty involved. (Watergate, covid lableak, etc.)

  we absolutely should be allowed to use the term conspiracy theory when warranted. User:Bluethricecreamman (Talk·Contribs) 23:44, 22 April 2026 (UTC)

• C. Pulling from other parts of the lead, could something like this work?:

  The dead Internet theory (DIT) is a theory that asserts that, since around 2016, the Internet has consisted primarily of bot activity and automated content manipulated by algorithmic curation. The DIT was originally proposed in a conspiracy theory alleging this to be part of a coordinated effort to manipulate and control the population and reduce genuine human interaction, however it has more recently^[when?] gained interest following the AI boom of the 2020s, with large language model (LLM) chatbots and text-to-image models emerging as technologies that could theoretically drown out human-authored content on the web.

  (Yes, I'm assuming "DIT" as a common acronym.) I hope this is able to detach the "the Internet is full of bots" concept from both the "it's a conspiracy!!!" and the "it's a genuine real phenomenon" components, allowing each interpretation to stand independently from the core concept. UniProbe62 (☎) 17:25, 1 May 2026 (UTC)

  The 2016 date is part of the conspiracy so should be removed from the first sentence. I overall agree with the goal of your proposal though. PK-WIKI (talk) 17:30, 1 May 2026 (UTC)

  I had been working on something similar:

  The dead Internet theory is a concept that asserts that much of the human activity on the Internet has been, or will be, taken over by bot activity and automated content promoted by algorithmic curation. The "dead" Internet consisting mainly of bots and algorithmic feeds is presented in contrast to the human-centered Internet of the early Information Age and the user-generated content posted to Web 2.0 social media networks.
  
  The dead Internet theory originated as a conspiracy theory claiming that, since around 2016, ...
  
  The theory gained renewed interest following the AI boom that began in the 2020s, with large language model (LLM) chatbots and text-to-image models emerging as technologies that could theoretically drown out human-authored content on the web.

  Something like the above, with a very generic first paragraph, second paragraph focused on the "original conspiracy", third paragraph focused on the AI boom bot reality. PK-WIKI (talk) 17:36, 1 May 2026 (UTC)

  I'm not entirely a fan of completely segregating the interpretations into their own paragraphs, as it means the paragraph immediately following the intro gets more focus, while the one after that gets sequestered into the bottom of the lead, almost like an afterthought. Irregardlessn't of how you order them, one of the interpretations would be a "main" one, and the other one would an "oh by the way" one. At least mentioning both in the intro paragraph means neither is immediately "preferred".

  Side note: I decided to phrase it as "The DIT is ___. The DIT was originally x, but is also now y" because doing that allows you to neutrally define what "DIT" is without immediately tying it to either "conspiracy" or "real phenonemon", and I included the interpretations in the same sentence in order to mention them at the same time and give them equal weight (something not possible with splitting them into paragraphs, or even into different sentences). "The DIT" is some noun, and "the conspiracy" and "the real phenomenon" are two distinct, related nouns (or concepts, but you get the idea). I can however see having a combined definition for the first paragraph, and then elaborating on each individually in the two subsequent paragraphs. UniProbe62 (☎) 18:25, 1 May 2026 (UTC)

  Again, we now have a source that discusses the emergence of this. The full DIT was and is a conspiracy theory, some sources focus only on the core arguments, which is described as the "leaner" version and contextualized within the full version. Per MOS:FIRST: "Do not overload the first sentence by describing everything notable about the subject." Trying to include both the original conspiracy theory and the new lean version in the first sentence is cramming everything about it into the first line, which results in a obfuscated prose that disguises the majority view on the full theory. We can, should, and do discuss spin off versions like the "strong," "weak," and "lean," however these have not replaced the original theory, and to understand them in full context, you need to start with the whole. Allowing the existence of the lean version to dilute the language that is used to define the theory as a whole is WP:PROFRINGE. GeogSage ^{(⚔Chat?⚔)} 20:05, 1 May 2026 (UTC)

  Okay, so the problem then is the definition of the basic term "DIT". Here, in the way I'm using it, I'm strictly defining it as "the theory that the Internet is full of bots". Just that one singular statement, alone. The "barest" definition of DIT possible, without any implication of a conspiracy or of it being a real phenomenon of the 2020s. I am not defining it as "the conspiracy", and would consider it inappropriate to use the term "DIT" to refer to the conspiracy interpretation as a whole. In other words, under my definition, "DIT" alone does not and should never also imply "conspiracy". That being said, it also doesn't necessarily imply "genuine phenomenon of the 2020s that is real and happening" (aka the scientific interpretation, or what we've so far been calling the "lean" version). It is simply the claim itself, nothing more. Maybe it would be best to give the conspiracy specifically the name "the DIT conspiracy", in order to avoid confusion with "bare" DIT, thus making "lean/scientific DIT" and "the DIT conspiracy" two instances of the separate overarching entity "bare DIT" (aka the fundamental claim shared by both interpretations).

  And yes, I know that me subjectively redefining DIT like this is... not even original research, just "original" on its own, which I imagine isn't entirely encyclopedia-friendly, but then again, apart from that one source, there are theoretically no extant reliable sources which don't exclusively define DIT as one or the other, or at least define it without also mentioning either "it's a conspiracy" or "it's real and it's happening". My "bare" definition is even leaner than the lean version, and, at least for the purposes of this article, is meant to imply nothing. The difference between scientific DIT and conspiracy DIT is just barely not enough to justify breaking out into two separate articles, which makes differentiating them kinda sorta really hard. So, the way I see it, there needs to be a bare definition in order for things to make sense. UniProbe62 (☎) 21:19, 1 May 2026 (UTC)

  Also, overload?? Really? The first sentence is a basic definition, and the second sentence establishes the two different, equally valid interpretations of said definition without going into excessive detail on either. UniProbe62 (☎) 21:27, 1 May 2026 (UTC)

  The DIT is not just the theory that the Internet is full of bots." The literature that discusses even the lite version, (Generative AI is Breathing New Life Into the Dead Internet Theory discussing the version put forward in Artificial influencers and the dead internet theory which is then referenced by several others) states it has four main claims: "1) Algorithms generate much of the Internet content 2) content influences perceptions and behaviors directed toward algorithmically-driven objectives 3) many Internet consumers have difficulty discerning between “real” and “fake” data (or, for that matter, “human-generated” and “AI-generated” data) 4) some GenAI byproducts (for example, deepfakes, AI Chat) create highly realistic yet fabricated content that undermine trust and propel misinformation." The full DIT is defined in the glossary of The Metaweb The Next Level of the Internet as "The Dead Internet Theory is a conspiracy theory that suggests the Internet has died and that much of the content we see online is now artificially generated by AI to manipulate the world population. The theory raises concerns about the impact of AI on propaganda, art, and journalism." We have other definitions of the full theory in the table I linked above that use similar language.

  These aren't two different interpretations that exist independent of each other. The DIT has two main parts, the lean version recognizes the core principles while stating, "Unfortunately, the conspiracy theorists augment these very plausible observations with their own mix of biases and agendas that lead to implausibility and absurdity, which in turn leads to rejection." Essentially, the conspiracy theorists have some good observations but their conclusion is implausible, we can still look at the observations though. This does not change the fact the DIT in its entirety is a conspiracy theory, just that, "Like all good conspiracy theories" the Dead Internet Theory is built on a "kernel of truth." The current state of the article seems in line with the literature, and the previous discrepancy seems to have been resolved by the new publication. GeogSage ^{(⚔Chat?⚔)} 23:05, 1 May 2026 (UTC)

  Mmmmmm... okay, 1. I was just using "the Internet is full of bots" as shorthand for the base claims, I didn't mean that literally, 2. I'd argue that, for the purposes of this article, "the full" Dead Internet Theory IS NOT the conspiracy. When we say "the full theory" or just "DIT", we should mean what I meant by "bare" DIT (I realize now differentiating between "bare" and "scientific" is moderately pointless). The name "DIT", on its own, should not include the conspiracy. Again, I propose that we just call it "the DIT conspiracy theory" in order to differentiate. Insisting that "the full" Dead Internet Theory includes the conspiracy and that "DIT" means the "full theory" is both useless and actively counterproductive, because doing so creates needless confusion that leads to, for example, you accusing me of WP:PROFRINGE, when that could have easily been solved by just defining a distinction between "Dead Internet Theory" and "DIT conspiracy", and sticking to that distinction throughout the article. For the purposes of this article, "DIT" cannot be used refer to the "full" version, as that fundamentally breaks both its actual definition and the entire article. And also this RfC. UniProbe62 (☎) 00:13, 2 May 2026 (UTC)

  The literature consistently defines the Dead Internet Theory as a conspiracy theory. Apparent discrepancies are reduced by recognizing a narrower variant that isolates a subset of its claims and stands in a part–whole relationship to the broader theory. However, this meronym should not be overemphasized relative to the holonym. GeogSage ^{(⚔Chat?⚔)} 01:12, 2 May 2026 (UTC)

  Okay, so then... what do we do?? How, exactly, are we supposed to distinguish between the two? What exactly do we even call it? "Not overemphasizing the meronym" means either a) we have to come up with a new term that isn't encyclopedically-unacceptable, as I tried to do with the subtly-tweaked "DIT conspiracy", or b) for half the article we simply talk about DIT without ever using the word "DIT". Madness. UniProbe62 (☎) 02:44, 2 May 2026 (UTC)

  We follow what the citations say, which is exactly what the article already is doing. The discrepancy we once had has now been resolved in an article discussing the various versions in the literature. We talk about the DIT, and then discuss the various versions that have spun out or been extracted from it. We already mention one of these in the lede (the lean version), the other two (strong and weak versions) were removed from the lede recently, which I disagree with but whatever. These are in the body of the article already. The DIT is a conspiracy theory, the lean version of it exists, and to understand the origins/history of it you first have to understand the conspiracy theory. GeogSage ^{(⚔Chat?⚔)} 23:43, 2 May 2026 (UTC)

  This seems much better than the current lead. I would support this proposal. Katzrockso (talk) 06:41, 2 May 2026 (UTC)

Discussion (conspiracy)

• I thought we'd already done this recently? What has changed? TarnishedPath^talk 07:26, 17 April 2026 (UTC)

  Nothing has changed. There are a few new sources that have come out, but most only tangentially touch on the topic, or are fairly low quality. There are a lot of predatory journals publishing on the topic, many making pretty extensive use of generative AI (ironic). There have been a few masters thesis's on the topic, and there is a lot of clickbait, but nothing has changed. GeogSage ^{(⚔Chat?⚔)} 07:31, 17 April 2026 (UTC)

  I'll have a look at BilledMammal's sources at a later time, but I think this is primarily a confusion about scope. Perhaps the topic of there being a metric fucktonne of generative AI content deserves its own article, but this isn't it. TarnishedPath^talk 07:43, 17 April 2026 (UTC)

  The source review is good for addressing discussion of Dead Internet Theory from 2025 onward and is absent any citations prior to 2025. It makes a good argument that the meaning has changed but doesn't speak to the relevance of prior meaning. Simonm223 (talk) 14:26, 17 April 2026 (UTC)

  It sounds to me, then, that these are really two different topics that happen to be called the same thing? If that is the case, they really should be split into two articles. If which one is primary is not easily determined, both can be disambiguated. Size wouldn't matter in this case, if these two things are actually two different subjects. I think there is at least one other similar such case on WP, with a "conspiracy theory" article and another that is just "theory", but I can't remember which articles. ~2026-23890-25 (talk) 04:04, 18 April 2026 (UTC)

  They are very closely related. We are comparing "some percentage of what is on the Internet was created by machines" with "some percentage of what is on the Internet was created by machines as a conspiracy to control our thoughts". A lot of issues apply to both meanings, such as

"Define 'Internet'. Does it include streaming? Software updates? Email?"

"Define 'percentage'. How do you compare a 1KB email to thousands of people with a 100MB video that is viewed twice?"

"Define "on the Internet". Do we count all the bytes that computers send to each other to establish connections?"

"Define 'Conspiracy' and 'mind control'. Are we talking about the CIA or the Illuminati controlling our deepest thoughts or are we talking about McDonald's trying to influence where we have lunch?"

--Guy Macon (talk) 04:31, 18 April 2026 (UTC)

Comment: In the original post discussing the theory, the "thesis" was "The U.S. government is engaging in an artificial intelligence-powered gaslighting of the entire world population." This is where the "conspiracy" part starts. There is at least one source from 2025 (the book Market-Oriented Disinformation Research) that describes a version of the DIT where society has already collapsed, and the some entity is maintaining the Internet to keep us ignorant of that. GeogSage ^{(⚔Chat?⚔)} 05:08, 20 April 2026 (UTC)

An arbitrary cut off of 2025 is ridiculous, and the sources reviewed are not vetted at all. Most of the sources listed are not going into extensive detail on the topic, and there are sources from 2025 that DO call it a conspiracy theory not included in the "review." Most of the articles that are using it as a hyperbole to underline a call to action or highlight some problems with AI are referencing the historic theory. Even if they are saying something like "used to be questionable, but new evidence makes it more likely," the concept of what it "used to be" provides the framework for what they are referencing. GeogSage ^{(⚔Chat?⚔)} 05:11, 18 April 2026 (UTC)

I certainly agree that the source review would be more useful if it extended back farther in time rather than just being 2025-6. Simonm223 (talk) 09:40, 18 April 2026 (UTC)

I have been working on that a bit with a table in my sandbox. I linked it in the discussion above, and I believe it was the inspiration for the one that scraped stuff since 2025. The one I've been working on has taken a while, largely because I'm trying to verify sources first. There is a lot of slop out there in the predatory journals that discusses the DIT. GeogSage ^{(⚔Chat?⚔)} 23:45, 18 April 2026 (UTC)

This is why determining academic consensus through source review like this isn't allowed per WP:RS. We cite source reviews, we don't create them. Thebiguglyalien (talk) 00:53, 20 April 2026 (UTC)

How do you figure that? How do we evaluate the WP:WEIGHT to give different claims if not by the WP:PROPORTION of sources? Katzrockso (talk) 01:20, 20 April 2026 (UTC)

If we want to say there's a general academic consensus for something, we have to cite a source saying there's general academic consensus: WP:RS/AC. Barring that, we say who supports what ideas without siding with anyone ourselves. Thebiguglyalien (talk) 01:39, 20 April 2026 (UTC)

My bad, my eyes skipped right over the part of your reply that said "academic consensus". Yes, I agree that stating an academic consensus requires an explicit source saying so. However, I don't think anyone here has made any claims that something is an academic consensus or suggested putting that in the main space. Katzrockso (talk) 03:33, 20 April 2026 (UTC)

• 2026 source (Berghel, Hal (2026). "Generative AI is Breathing New Life Into the Dead Internet Theory". Computer. 59 (1): 132–139. doi:10.1109/MC.2025.3616665. Retrieved 20 April 2026.) finally explicitly discusses different versions. Have included a section on "leaner" version using this under "claims," added bit to lede 3rd paragraph that was already discussing this divide without the explicit mention, and the expert views section. Glad to say the 3rd paragraph is no longer boarder line WP:Synth in my opinion. Probably needs reworking, but thought I'd get that source incorporated ASAP.

GeogSage ^{(⚔Chat?⚔)} 06:36, 20 April 2026 (UTC)

• Is this an actually conspiracy theory that people believe in or more of a cultural discourse thing? I read the enwiki article years ago and the latter was what it was presented as and what I believe still stands true. Gotitbro (talk) 10:16, 19 April 2026 (UTC)