“The Dead Internet Theory is an online concept which asserts that the internet has become largely devoid of authentic human activity, having been overtaken by bots, automated algorithms and artificial intelligence. Originating as a fringe conspiracy theory claiming a coordinated, artificial takeover occurred around 2016, it has evolved into a serious critique of the modern web.” – The Dead Internet – Online theory
Concerns about whether online life is still genuinely human arise from the way commercial incentives, recommendation engines and synthetic content combine to structure what most people now see. Instead of an open, messy patchwork of individually maintained sites and conversations, users encounter platform feeds tuned for engagement, advertising systems optimised for click-through, and content factories pumping out low-cost material, increasingly generated or assisted by artificial intelligence. At the same time, measured bot traffic has grown to rival or exceed human traffic on the public web, with security firms repeatedly reporting that automated requests constitute around half or more of all observed activity 1,2,11. This dual evolution of commercial curation and machine traffic underpins both the plausibility and the anxiety of the claim that the internet has somehow become hollowed out.
Substance of the theory: from conspiracy to structural critique
In its strong, conspiratorial form, the dead internet idea asserts that a coordinated project, often attributed to states or large corporations, replaced most human interaction online with bots and synthetic content sometime around 2016 or 2017 1,6,9. Proponents argue that what users experience as spontaneous conversation, trending topics or grassroots movements is in fact orchestrated output from networks of automated accounts and recommendation systems, designed to steer opinion, sell products or maintain social control. This narrative typically includes the claim that whole platforms are now dominated by AI-written posts, automatically generated comments and search results padded with machine-authored pages, while authentic forums have been quietly deprecated or buried by ranking algorithms. More moderate versions strip out the hidden-conspiracy element and focus instead on measurable trends: the growth of bot traffic, the homogenisation of feeds and the industrialisation of content production as reasons to say that the web no longer functions as a primarily human commons 1,7,12.
Empirically, the theory draws on several strands of evidence. One is the long-run increase in automated traffic measured by cybersecurity and infrastructure firms. Imperva, for example, reported in 2016 that bots accounted for 52% of observed web traffic, a finding that has been repeatedly cited by commentators as proof that non-human requests have overtaken human browsing 1,8,15. More recent data suggest that bots consistently represent around half of total traffic, with so-called bad bots alone accounting for about 30% to 37% of requests in the mid-2020s 2,3,5,11. A second strand involves the documented role of bots and coordinated inauthentic behaviour in political manipulation, particularly around the 2016 US election and subsequent campaigns, where state-linked and commercial actors used automated accounts to amplify polarising content. A third strand concerns the rapid rise of generative AI systems capable of producing text, images and video at scale, lowering the cost of flooding search results, review sites and social platforms with synthetic material 7,8,14.
Origins and evolution in online subcultures
Although the sense that the web has become fake or empty circulated informally on image boards and small forums for years, the term in its current form is widely traced to a 2021 manifesto posted on Agora Roads Macintosh Cafe by an anonymous user writing as IlluminatiPirate 1,7,9,12,14,15. That post synthesised earlier discussions from communities such as Wizardchan and other fringe boards, where participants had complained since roughly 2016 that search results felt sanitised, independent blogs had disappeared from the first pages of major engines, and social feeds showed the same kinds of viral content across different platforms. Academic work on online subcultures notes that the precise genealogy of the idea is hard to reconstruct, partly because these spaces are transient and poorly archived 1,7. However, researchers broadly agree that the theory crystallised in the late 2010s and early 2020s amid wider disillusionment with platform capitalism, targeted advertising and pervasive surveillance. The concept subsequently moved from esoteric forums into mainstream awareness through coverage in outlets such as The Atlantic, which framed it as a striking metaphor for post-2016 discontent with the changing feel of the internet 1,7,9,12.
Mathematical framing: bots, traffic shares and content dominance
Where the discussion touches quantitative analysis, it is useful to distinguish between traffic shares, content shares and attention shares. Traffic is often decomposed into human and bot components, with security firms estimating proportions over time. If T denotes total observed web requests over a period and B denotes requests generated by bots, then the bot share is s_B = B / T. Reports such as Impervas indicate that s_B has hovered around or above 0,5 in several recent years, and that within B the subset of malicious or bad bots accounts for a further share s_{BB} of perhaps 0,3 to 0,37 of T 1,2,11,15. However, a high s_B does not by itself imply that most content is machine-generated, as many bots are crawlers that index existing human-authored pages. To move from traffic to content, one would in principle need measures of the number of distinct documents created by humans versus AI systems, or the proportion of tokens on the web attributable to each source, quantities that are much harder to observe directly. The theory effectively claims that both the content share c_{AI} of AI-generated material and the attention share a_{AI} it captures have risen towards 1 in key domains such as search, reviews and low-value news.
A more formal sceptical perspective models online environments as mixtures of human and automated agents interacting strategically. If H_t represents the density of active human accounts at time t on a given platform and A_t represents automated or AI-driven accounts, then one can consider the ratio R_t = A_t / (H_t + A_t) as an index of automation. The conspiracy version of the theory implicitly asserts a structural break: that there exists some t_0 around 2016 for which R_t jumps discontinuously to a value close to 1 across most major platforms. Available empirical studies instead tend to show gradual, heterogeneous increases in automation, with spikes around politicised events but no universal step change 1,11,14,15. In other words, the mathematical claim required to sustain the idea of a sudden death is not supported by measured time-series, although local environments such as thinly moderated forums or comment sections can indeed become dominated by bots.
Parameters, actors and mechanisms
Analytically, the situation can be decomposed into several interacting parameters. First, there is bot sophistication, including the ability to mimic linguistic style, temporal posting patterns and network structures of human users. Advances in generative models and reinforcement learning from human feedback have substantially increased this capability, which raises the effective detection cost for platforms and security firms 3,7,14. Second, there is the economic incentive to deploy bots, whether for advertising fraud, search-engine optimisation, political influence or spam. As long as the expected gain from automation, weighted by the probability of evasion, exceeds the cost of creation and maintenance, rational actors will continue to add synthetic agents to the system. Third, there is platform design, particularly the extent to which ranking and recommendation algorithms amplify content based on engagement metrics that do not discriminate between human and automated signals. When engagement-driven optimisation is combined with cheap synthetic content, feedback loops can arise in which low-value or repetitive material is preferentially surfaced, degrading perceived authenticity even when a majority of underlying accounts remain human 1,14,15.
Schools of thought: literal takeover versus metaphorical death
Current commentary on the dead internet idea falls roughly into three camps. Literalists within conspiracy communities assert that the majority of visible online activity is now artificial, and that this shift is the result of coordinated decisions by states, intelligence agencies or major corporations to replace unpredictable human discourse with controllable simulations 1,9,14,15. A second camp treats the concept as a dark metaphor for platform consolidation, arguing that even if most users are still real people, the structure of attention has become so dominated by a small set of intermediaries, optimised for advertising and surveillance, that the early promise of a diverse, user-driven network has effectively died 1,7,12. A third camp, typically comprising researchers and journalists, sees value in the theory as a diagnostic of public unease about bots, AI and information quality, while rejecting its empirical overreach. From this standpoint, the important questions concern how synthetic activity interacts with human behaviour to shape public discourse, not whether some threshold of machine dominance has technically been crossed 7,11,14.
Critiques, tensions and empirical limits
Critics raise several key objections. First, they point out that much of the cited evidence, such as the Imperva traffic figures, predates the supposed 2016 death and reflects longstanding infrastructural bot activity rather than a sudden, concealed replacement of humans 1,8,15. Second, they emphasise that while bots play documented roles in spreading misinformation and manipulating metrics, numerous empirical studies of social media networks still find dense clusters of human interaction, creative communities and grassroots organising that cannot be reduced to synthetic output 7,14. Third, the conspiracy framing tends to underplay how ordinary users themselves participate in homogenisation by converging on the same platforms, formats and templates, and by optimising their own content for algorithmic visibility. There is also a methodological tension: companies holding the most detailed data about bot prevalence and AI-generated content have limited incentives to release that data publicly, making definitive measurement difficult and leaving space for speculative narratives to flourish 1,11,13.
Why the idea continues to matter
Despite its evidential weaknesses in the strong form, the dead internet concept continues to matter because it condenses widespread anxieties about authenticity, agency and information quality in an AI-saturated media environment. Even if human users still generate a large share of total content, the knowledge that bots may represent half of all traffic and that generative systems can fabricate convincing text, images and interactions at scale erodes trust in what is seen and read online 1,2,3,11. This loss of trust has practical consequences for journalism, education, democratic deliberation and everyday social life, as individuals struggle to distinguish genuine peers from synthetic personas, and verified reporting from machine-spun spam. The theory also draws attention to structural power: whoever controls recommendation algorithms, search rankings and AI training data effectively shapes the informational environment in which most people now live. In that sense, debates about whether the internet is literally dead are less important than the concrete policy and design choices around transparency, auditability, bot labelling and support for independent, human-scale online spaces. The persistence of the idea serves as a warning that without credible, shared metrics and governance, perceptions of a hollowed-out web will continue to grow, regardless of the underlying statistics.
References
1. Dead Internet theory – Wikipedia – 2023-03-06 – https://en.wikipedia.org/wiki/Dead_Internet_theory
2. Bots Now Make Up Nearly Half of All Internet Traffic Globally | Nasdaq – 2024-04-16 – https://www.nasdaq.com/press-release/bots-now-make-up-nearly-half-of-all-internet-traffic-globally-2024-04-16
3. AI-driven bad bots account for 37% of internet traffic in 2024 … – 2026-02-02 – https://journalrecord.com/2026/02/02/ai-bad-bots-internet-traffic-2024/
4. Bots now make up the majority of all internet traffic – The Independent – 2025-04-15 – https://www.the-independent.com/tech/bots-internet-traffic-ai-chatgpt-b2733450.html
5. Human and bot web traffic share 2024 – Statista – 2025-07-17 – https://www.statista.com/statistics/1264226/human-and-bot-web-traffic-share/
6. Dead Internet Theory – Wikipedia – 2024-07-01 – https://de.wikipedia.org/wiki/Dead_Internet_Theory
7. Between the Self and Signal: The Dead Internet & a Crisis of … – https://openresearch.ocadu.ca/id/eprint/4676/1/Between%20the%20Self%20and%20Signal%20__%20The%20Dead%20Internet%20&%20a%20Crisis%20of%20Perception.pdf
8. [PDF] The Dead Internet Theory: Investigating the Rise of AI-Generated … – https://www.rivista.ai/wp-content/uploads/2025/05/Dead_Internet_Theory.pdf
9. ?????? ???????? ????????? – ????????? – 2024-08-05 – https://ru.wikipedia.org/wiki/%D0%A2%D0%B5%D0%BE%D1%80%D0%B8%D1%8F_%D0%BC%D1%91%D1%80%D1%82%D0%B2%D0%BE%D0%B3%D0%BE_%D0%B8%D0%BD%D1%82%D0%B5%D1%80%D0%BD%D0%B5%D1%82%D0%B0
10. Bots Drive 30% of Global Web Traffic, Outpacing Human Users in … – 2025-07-02 – https://cyberpress.org/bots-drive-30-of-global-web-traffic/
11. Chart: Are Bad Bots Taking Over the Web? – 2026-03-10 – https://www.statista.com/chart/amp/35950/share-of-global-web-traffic-generated-by-humans-and-bots-since-2018/
12. Théorie de l’Internet mort – Wikipédia – 2023-06-12 – https://fr.wikipedia.org/wiki/Th%C3%A9orie_de_l’Internet_mort
13. Bots Compose 42% of Overall Web Traffic; Nearly Two-Thirds Are … – 2024-06-25 – https://www.akamai.com/newsroom/press-release/bots-compose-42-percent-of-web-traffic-nearly-two-thirds-are-malicious
14. The Dead Internet Theory: Is Most of the Web Just Bots and … – 2025-10-31 – https://thisweekinsciencenews.com/blog/2025/10/31/the-dead-internet-theory-is-most-of-the-web-just-bots-and-fake-content/
15. An Investigation into the Dead Internet Theory | New Space … – 2025-09-05 – https://newspaceeconomy.ca/2025/09/05/an-investigation-into-the-dead-internet-theory/
