Origin — The Ur-Text and the Manifesto

The Dead Internet Theory does not have a single clean moment of origin. Early murmurings — discussions of the internet feeling hollow, repetitive, and increasingly non-human — appeared on the imageboard Wizardchan and on 4chan’s paranormal board at some point before 2021. The exact posts have not been recovered or definitively dated; they exist in the theory’s mythology rather than its documented record.

What is documented is the text that gave the theory its name, its structure, and its first significant audience. In early 2021, a user operating under the handle “IlluminatiPirate” posted a lengthy thread titled “Dead Internet Theory: Most of the Internet Is Fake” on Agora Road’s Macintosh Café — a small, niche forum associated with retrocomputing aesthetics and fringe internet culture. The post explicitly acknowledged building on earlier discussions from Wizardchan and 4chan, but it synthesized those threads into something coherent enough to travel. Journalist Kaitlyn Tiffany, writing for The Atlantic in September 2021, called it “the theory’s ur-text.”

IlluminatiPirate’s argument was structured around a felt experience rather than hard data: the internet of the mid-2010s felt different from the internet of the early 2010s. The diversity of voices, the emergence of unexpected content, the sense of stumbling across genuine human creativity — all of it seemed to have been replaced by something homogeneous, repetitive, and oddly purposeful. The explanation he offered: the internet had been deliberately killed. Since approximately 2016 or early 2017, he argued, the majority of internet content had been replaced by AI-generated output and bot activity, coordinated by a combination of government agencies and large corporations. The old internet — messy, human, alive — was effectively gone. What remained was a simulation of it.

The post gained traction quickly. It was shared across Reddit — including on a Joe Rogan fan subreddit with hundreds of thousands of members — and spread to tech forums including Hacker News. YouTube creators produced explainer videos; a Spanish-language summary alone garnered over a quarter of a million views. By September 2021, when The Atlantic published its coverage, the theory had traveled from a fringe forum into mainstream discourse — before the emergence of ChatGPT, before the AI content explosion of 2023, and before the term “AI slop” had entered the vocabulary.

---

Section I — The Theory’s Two-Component Structure

Analysts and researchers who have examined the Dead Internet Theory consistently identify it as having two separable components, which are often conflated in popular coverage but which have very different evidential statuses.

---

Section II — The Data: What Bot Traffic Measurements Actually Show

The most authoritative longitudinal dataset on automated internet traffic is Imperva’s annual Bad Bot Report, now in its 12th edition (2025). Its methodology covers billions of web requests across thousands of domains globally and classifies traffic by source — human, good bot (search crawlers, monitoring tools), and bad bot (scrapers, account takeover tools, content generators, engagement manipulators). The trend the data reveals is unambiguous.

The 2025 Imperva report — based on analysis of 13 trillion blocked bad bot requests across thousands of domains — found that for the first time in a decade, automated traffic exceeded human traffic, accounting for 51% of all web activity in 2024. Bad bot activity alone has risen for six consecutive years, with malicious bots now accounting for 37% of all traffic. Imperva’s report attributes the acceleration primarily to the widespread adoption of AI and large language models, which have dramatically lowered the barrier to bot creation: users who previously needed programming skills to deploy a bot can now do so using off-the-shelf tools or simple natural-language prompts.

---

Section III — Enshittification: The Parallel Framework

Running parallel to the Dead Internet Theory — and providing its most credible mainstream equivalent — is the concept of enshittification, coined by Canadian journalist and activist Cory Doctorow in November 2022. Selected as the American Dialect Society’s 2023 Word of the Year and Australia’s Macquarie Dictionary’s 2024 Word of the Year, enshittification describes the predictable cycle by which digital platforms decay.

Doctorow’s model has three stages. First, a platform provides genuine value to users to achieve scale and lock-in. Second, once users are dependent, the platform begins extracting value from them to benefit business customers (advertisers, sellers, publishers). Third, once business customers are also locked in, the platform extracts value from them too, ultimately serving only shareholders. Facebook’s trajectory from an ad-free social network to a feed dominated by paid promotion; Amazon’s search results where the first page is predominantly paid placements and the best-priced item typically appears 20 spots down; Google Search increasingly returning SEO-optimized content farms rather than useful pages — these are Doctorow’s primary examples.

Where the Dead Internet Theory blames a coordinated external conspiracy, enshittification offers a structural explanation: the same outcome — a less human, less authentic, more commercialized internet — emerges from the internal logic of platform monopoly under conditions of weakened regulation, captured antitrust enforcement, neutralized worker power, and eliminated competitive pressure. No single actor needs to have planned it. The architecture of the incentives produces it inevitably.

---

Section IV — The AI Content Flood: Accelerant Since 2022

IlluminatiPirate’s 2021 manifesto was written before the release of ChatGPT (November 2022), before the mass adoption of image-generation models, and before “AI slop” became a recognized cultural phenomenon. This timing is one of the more striking aspects of the theory’s history: its core prediction — that AI-generated content would displace human-authored material at scale — has been substantially validated by developments that postdate the original claim.

Since late 2022, AI-generated text, images, and video have proliferated across virtually every major platform. TikTok feeds have filled with AI-narrated videos using stock footage and algorithmically optimized scripts. Facebook and Instagram feeds circulate AI-generated images — particularly of sentimental subjects, natural disasters, and religious imagery — that receive millions of engagements from real human accounts. Content farms using AI tools produce thousands of articles per day for SEO purposes, saturating search results. Pinterest, Reddit, and even LinkedIn have reported significant increases in AI-generated content. The question is no longer whether AI content is present at scale on major platforms, but what proportion of what any given user encounters in a typical session is machine-made.

The phenomenon has acquired its own vocabulary. “AI slop” refers to low-quality AI-generated content optimized for engagement rather than quality. “Slop farms” are operations producing this content at industrial scale. A 2025 academic survey paper on arxiv (A Survey on Artificial Interactions) specifically examined the intersection of AI-generated content and the Dead Internet Theory framework, finding sufficient structural overlap to treat the theory as a meaningful lens for analyzing contemporary internet dynamics.

---

Timeline — Key Moments in the Theory’s History

---

Section V — Responses and Reception

The Dead Internet Theory has attracted a notably wide range of responses — from flat dismissal to substantive partial endorsement — that tend to correlate with which component of the theory is being evaluated.

Mainstream media reception has been mixed but increasingly willing to engage with the technical component. The Atlantic‘s 2021 piece called it “wrong but feels true,” a formulation that captures the dominant mainstream position: the coordinated conspiracy is rejected, the underlying phenomenology of a degraded internet is taken seriously. The Guardian, BBC, Forbes, and others have covered it in terms that treat the technical observation as newsworthy even while distancing from the conspiratorial framing.

Academic engagement has been sparse but growing. A paper published via Springer in 2023 examined “artificial influencers and the dead internet theory” as a legitimate frame for analyzing social media authenticity. The 2025 ArXiv survey paper takes the theory seriously as a descriptive framework for AI-content proliferation. A 2023 CRC Press volume on digital culture includes dedicated discussion of it. The academic consensus, where one exists, treats Component 1 as increasingly empirically supported and Component 2 as unfalsifiable speculation.

Counter-arguments focus on several points: that human traffic remains enormous in absolute terms even if declining as a proportion; that most bot activity is economically motivated (advertising fraud, scraping, spam) rather than politically motivated; that the experience of the internet feeling “dead” is partly an artifact of social media’s algorithmic curation rather than an actual absence of humans; and that nostalgia for the “old internet” romanticizes a web that was also full of spam, scammers, and low-quality content.