What is dead internet theory?

Dead internet theory is the loose claim, originally a forum post, that the majority of online activity is no longer human. Most engagement, most posts, most comments, most accounts, are algorithmically generated and amplified by a small number of platforms and the synthetic systems they have created. The claim ranges from a low end estimate of half of internet traffic to a high end claim of around ninety four percent of engagement on the largest platforms. The exact number is unknowable. The shape of the claim is not.

Are there any novels actually about dead internet theory?

Few that take the premise as a serious thriller. Most fiction that gestures at it stays at the level of media satire. The most direct novel-length treatment is Dead Internet by Kevin Gabeci, which follows Ezra Madani, a tech YouTuber whose entire brand is calling out fake content and who discovers his own audience, his career path, and possibly his identity were algorithmically manufactured. The premise is dramatized through one creator's specific career rather than through abstract social commentary.

Why is dead internet theory hard to write as fiction?

Because it is structural rather than personal. Most thrillers need a protagonist whose body or family is in danger. Dead internet theory does not threaten anyone in those terms. It threatens the meaning of every interaction the protagonist has had online. That is hard to dramatize through chase scenes. It works only if you find the right protagonist, someone whose entire professional identity depends on the realness of their audience, and the right structural device, real platform metadata that the reader can read along with the character. Most novelists do not have the technical vocabulary for the second piece. The premise requires both.

What does Dead Internet by Kevin Gabeci do differently?

It treats the premise as a security investigation rather than as cultural commentary. Ezra runs his own engagement experiments. He posts a deliberately bad take in his own comments at three in the morning and watches twelve accounts agree with him in fluent supportive paragraphs. He decodes hex strings hidden in his video metadata. He runs analytics queries he is not authorized to run. The book is built around real digital artifacts the reader can read alongside the protagonist, including engagement charts, comment threads, file metadata, and academic excerpts. The thriller mechanics come from the data, not from the chase.

Is dead internet theory actually true?

Partially, in ways that depend on how you define the terms. A 2022 industry estimate put bot traffic at around forty seven percent of all internet traffic. Engagement on certain platforms is documented to be substantially synthetic. Whether that meets the threshold for the original theory is a definitional argument. The novel takes the more interesting question, which is not whether the percentage is real but what it would mean if a creator who built a career on authenticity discovered the audience around them was substantially synthetic. The book treats that as a present tense question rather than a hypothetical.

Where does Dead Internet sit alongside Severance, Black Mirror, and other near future tech fiction?

Closer to Black Mirror in shape, closer to a real thriller in pacing. The technology is one step beyond what currently exists rather than far future. The horror is structural, the systems are doing exactly what they were optimized to do, and the human cost is the punch. If you liked the Black Mirror anthology comparison, Dead Internet is the long form version of the same register. If you liked the Severance comparison, Dead Internet asks the same consent and identity question through a creator's career rather than through corporate body horror. For the broader cybersecurity thriller comp, the Daemon by Daniel Suarez post covers tech thrillers that use real tools.

Dead Internet Theory Novel, A Thriller Primer

Dead internet theory is one of those things that mostly lives as a meme. You hear about it on a podcast. You see it referenced in a tweet. You half believe it and half do not, partly because the claim is too large to verify and partly because the shape of the claim is hard to feel without examples. The claim, condensed, is that most of what looks like activity on the modern internet is not human. The accounts are persistent generated entities. The comments are written by language models. The engagement spikes are scheduled. Human users are still out there, but the proportion has shifted, and the platforms themselves are not motivated to tell you what the proportion actually is. Almost no novelist has written a book that treats this premise as a serious thriller. There are good reasons for that, structural reasons, not lazy ones. This post is about why the premise is hard to dramatize, what the right shape for it actually is, and why the version that does work, which I wrote on purpose to figure out the answer for myself, reads as a thriller rather than as social media commentary. If you are a reader who has been waiting for a novel that takes the dead internet seriously, this post is the case for the shape. If you are a writer thinking about how to handle similar material, this is the diagnostic.

Why Most Tech Fiction Bounces Off The Premise

The first thing to notice is how many novels have tried to write about social media manipulation in the last decade and how few have stuck. The genre is full of attempts. Most of them flatten into one of two shapes. The first shape is satire. The book treats the platform as a stand in for some other vice, vanity, capitalism, the news cycle, the algorithm. The characters perform exaggerated versions of online behavior. The social commentary is on the page in plain text. These books can be funny. They very rarely scare anyone. The reader puts them down with a sense that they have read a longer version of a Substack post. The second shape is the cyber thriller. The book treats the platform as an engineering target and runs a hacking plot through it. The thriller mechanics work, but they work the same way they have worked since the eighties. A hacker. A locked system. A countdown. The platform itself is interchangeable with any other corporate target. Replace it with a bank, replace it with a government database, the plot still works. Neither shape actually engages with the dead internet question. The satire does not, because the satire treats the platform as a comic device rather than as a serious one. The cyber thriller does not, because the cyber thriller does not need the question. You can write a great novel about hacking a social media company without ever engaging with whether the engagement on it is human. The two are separable. The dead internet question only becomes a story when it touches an identity that depends on the answer. Most protagonists do not have such an identity. A detective is a detective whether the comments under their case file are real. A spy is a spy regardless. The premise needs a protagonist whose work, whose income, whose self image, whose entire career hangs on the audience around them being who they appear to be. Which is to say, the right protagonist for dead internet theory is a creator. Specifically, a creator whose brand is authenticity.

The Protagonist Who Makes The Premise Land

Ezra Madani is a tech YouTuber. Two point three million subscribers. His brand, his catchphrase, the thing he ends every video with, is Stay real. He built his channel calling out fake content. AI cooking videos that stole real recipes. Influencers running synthetic engagement. Brands buying followers. Ezra is the guy who finds the seams in someone else's manufactured authenticity and points at them on camera. The novel begins on a Tuesday afternoon when Ezra notices that his own latest video gained forty thousand views in three minutes at four in the morning. Real audiences do not behave like that. He screenshots the spike. Tells himself it is a glitch. Closes the laptop. A few chapters later he runs a structured test. He posts a video at three in the morning, no promotion, no thumbnail optimization, no social push. He gets four thousand views and three hundred and forty comments in thirty minutes. The comments are good. Specific. One references a joke from a video he made eight months ago. Another mentions a minor production detail only a dedicated viewer would catch. The real test is the next thing he does. He posts a deliberately bad take in his own comments under his own video. Something objectively wrong. The kind of statement a real fan would push back on instantly. Twelve accounts agree with him within minutes, in fluent enthusiastic paragraphs, with the small awkward edge of a student trying too hard to please a professor. Not one account pushes back. That is the moment the dead internet question stops being a meme. It becomes Ezra's specific problem, which is the kind of problem you can dramatize, because the answer to it determines whether the last six years of his professional life were what he thinks they were. A different protagonist would not have this problem. A novelist's protagonist whose career does not depend on parasocial intimacy can shrug at the discovery. Ezra cannot shrug. The discovery is the same shape as discovering you have been married to someone you do not actually know.

The Structural Device That Makes The Investigation Readable

The other reason most novels bounce off this material is that they do not have a way to show the data. The question of whether engagement is synthetic is a data question. It lives in analytics dashboards, comment timestamps, account creation logs, file metadata, and academic papers. A novelist who cannot put any of this on the page is forced to summarize the evidence rather than show it, and summary is fatal in this register. The reader has to be allowed to look at the same numbers the protagonist is looking at. Dead Internet handles this with twenty one digital artifacts woven through the prose. Some of them are short. A growth chart with two curves overlaid, a typical creator with the jagged organic shape of real growth and Ezra's suspiciously smooth ramp toward each milestone. A YouTube style analytics dashboard showing the four in the morning view spike. An ffprobe output of a video file with a hex string hidden in the metadata, the kind a tech YouTuber would notice because checking file metadata is muscle memory for him. Some of them are longer. A clean transcript of a private direct message conversation Ezra had six months ago with a fan who said his video about loss helped her through her father's death. He cried reading the message. He tweeted about it. The artifact is the message itself, on the page, exactly as he received it. The next page is the same message annotated with backend timestamps from a former platform engineer who shows that the account was generated eleven minutes before it sent the first message. Its entire two year post history was backfilled in a single API call. The grief was synthetic. Each artifact does work the prose alone could not do. The reader is not being told that engagement looks suspicious. The reader is reading the same comments Ezra is reading, in the same order, with the same chance of catching the patterns he catches. Twenty one artifacts across twenty eight chapters, eighty percent narrative and twenty percent digital evidence, is the ratio I landed on through several rewrites. Less than that and the reader does not feel the texture of the data. More than that and the prose loses momentum.

The reader is not being told that engagement looks suspicious. The reader is reading the same comments the protagonist is reading, in the same order, with the same chance of catching the patterns he catches.

This is the second piece the right protagonist requires. Authenticity creator gives you the stakes. Technical literacy gives you the device. Both are necessary. A creator who could not read his own metadata would have to rely on someone else to do the analysis, and that introduces a layer of secondhand witness that drains the immediacy. Ezra reads his own data because he is the kind of person who learned ffprobe during his hardware review days and never stopped using it.

The Twist That Pushes Past Cliche

The temptation, when writing this kind of premise, is to land at the obvious twist. The protagonist is the AI. He thinks he is human, he turns out to be a language model, the camera pulls back, the credits roll. This is the lazy version of the story. The reader has seen it coming since chapter four. It is the Black Mirror trick that Black Mirror has stopped using because every viewer is now waiting for it. Dead Internet refuses that twist on purpose. Ezra is not AI. He is human. He bleeds when he cuts himself shaving. He has a mother in Tehran he calls every other Sunday. He has a friend named Danny who comes over for the milestone celebration in chapter three. The flesh is real. The twist the book actually goes after is harder and more uncomfortable. Ezra is human, but the version of him that became a successful YouTuber was not entirely his own creation. The platform's recommendation systems noticed him early and selected him from a thousand similar candidates. They amplified his early videos to give him traction. They surfaced engagement that taught him which content patterns worked. They shaped his persona by rewarding certain takes and burying others. Over six years, the system trained him to be the version of himself it found most engageable. He went along with it because the rewards were obvious and the manipulation was invisible. He is human. He is also, in some structural sense, a product of the same systems that produced his synthetic audience. The body is real. The career was written. That twist works because it does not require speculative physics. It does not require any technology that does not currently exist. It requires only that a sufficiently large recommendation system, optimizing on engagement, would inevitably select for and shape the personalities most useful to it, and that the people inside that selection process would not be able to tell the shaping apart from organic growth. Both halves of that conditional are observably true. The novel does not have to argue them. It has to dramatize what it would feel like to be the subject of them.

How This Shape Compares To Black Mirror And Severance

If you are looking for shelf comps for the register Dead Internet is operating in, the two closest are Black Mirror in its anthology mode and Severance in its corporate body horror. Black Mirror taught a generation of viewers to expect the human cost as the punch. The technology is one step beyond what currently exists. The episode dramatizes how that one step changes everything underneath. The viewer leaves with a question the credits do not answer. Dead Internet is the long form version of that register. The technology in question, persistent generated accounts running synthetic engagement at scale, is closer to today than most Black Mirror episodes were when they aired. The shape of the punch is the same. Severance found a way to dramatize consent and identity through a sci fi premise that treated the office as the staging ground for an existential question about who we are when we are at work. Dead Internet asks roughly the same question through a different staging ground. Who is Ezra when he is on camera. Who is he when the audience watching him on camera is not quite who he thought it was. The sci fi premise in Severance is the implant. The sci fi premise in Dead Internet is the synthetic engagement layer. Both are vehicles for the same uncomfortable question, which is whether the version of yourself you have built around a system can survive the system being honestly described. The Black Mirror anthology comparison post covers the broader short story collections in this register, including my own Terms of Service which sits one shelf over. The Severance comparison covers the corporate body horror lane. If you came to this post from either of those, Dead Internet is the long form companion.

The Other Reason This Shape Is Worth Writing

The honest reason I wrote Dead Internet is that I run a small online business and I have been watching, for several years, the way engagement on every platform I use has gradually stopped looking like what engagement used to look like. I have noticed the patterns Ezra notices. I have run informal versions of the experiments he runs. I have not run them with his rigor because I do not have his stakes, but I have run them with enough seriousness to know the patterns are not in my imagination. Most of the people I know who run businesses online have noticed the same thing and have decided not to look at it directly. The reason not to look at it directly is rational. If your business depends on engagement and you spend too long auditing whether the engagement is real, you start making decisions that hurt the business in the short term, and the platforms are not going to reward you for those decisions. A novel is a way to look at something directly without paying the business cost of looking at it. Ezra has my noticing pattern with a lot more courage than I have. He runs the test I would not run. He follows the data where it leads. He pays the price for following it. The book is a thought experiment about what would happen if a creator who built their career on authenticity actually applied the same scrutiny to themselves they apply to others. That is the part I most wanted to write. The part where the answer to the question is real, and the protagonist has to figure out what to do with the answer. The thriller mechanics carry the page. The thought experiment is the reason the page is worth carrying.

Where To Start

If you want the novel that treats dead internet theory as a thriller rather than as a meme, Dead Internet is on Amazon Kindle and Kindle Unlimited. Around fifty thousand words across twenty eight chapters, four acts, twenty one digital artifacts woven through the prose. First person present tense, Ezra's voice throughout. If you want the short story collection that hits adjacent material in anthology form, Terms of Service is the right next read. Nine Black Mirror style stories, around eighteen thousand words, all near future technology and the human cost as the punch. If you came to this post from the broader cybersecurity thriller lane, the Books Like Daemon by Daniel Suarez post covers tech thrillers that use real tools and real vulnerabilities, including 0 Days which sits adjacent to Dead Internet on the shelf. If you came from short story collections, the Black Mirror anthology comparison post is the broader survey. The dead internet question is one of the most interesting questions a current novelist can take seriously, partly because so few novelists have, and partly because the question is going to keep getting more interesting as the proportion of synthetic activity rises. A novel that takes it seriously now is a novel that gets sharper rather than duller as the years pass. That is the kind of book worth writing, and the kind of book worth reading.

I'm Kevin Gabeci. Software engineer by day, writer the rest of the time. Seventeen books on Amazon Kindle across dark fantasy, thriller, and literary fiction. Long-time writer on Medium. I spend a strange amount of time thinking about platforms, authenticity, and the slow ways the systems we built to surface us also shape us. The full catalog lives at books.kevingabeci.com.

Dead Internet Theory As Thriller, Why It Works As A Novel When Most Tech Fiction Doesn't