Viral online culture is now largely fake, secretly manufactured by AI influencers, synthetic personas, and coordinated engagement

The case people make

Take the strong version seriously, because unlike most entries in this encyclopedia its believers can point to named, dated, documented exhibits rather than rumor. The case builds in layers, and each layer is real on its own terms.

Synthetic personas already work.If an openly artificial influencer can accumulate a real audience and real brand income, and one demonstrably can, then the claim that a hidden one could do the same is not far-fetched. In 2025 an AI “actress” drew a wave of coverage and industry alarm simply by seeking representation, which suggested to many that synthetic figures are being pushed into culture whether or not audiences asked for them.

The volume is overwhelming. Independent estimates in 2025 put AI-generated material at roughly half of new English-language web articles; streaming services reported removing tens of millions of spam tracks; a notable share of the fastest-growing video channels were AI-made. When that much of what surfaces is synthetic, the intuition that a viral moment might be manufactured stops looking paranoid.

And coordination is not hypothetical. AI firms have themselves disclosed shutting down operations that used generative models to write fake comments, invent biographies for fake accounts, and seed propaganda, targeting real political audiences. That is the exact mechanism the theory describes, confirmed by the companies whose tools were used.

The believer's strongest move is not invention; it is extrapolation. Real synthetic influencers, real slop at scale, real coordinated bot operations. If all of that is documented, why assume the visible cases are the only ones?

That is the honest core of the suspicion: not that any specific viral moment has been proven fake, but that the ingredients of large-scale fabrication plainly exist, are cheap, and are already in use, so treating online popularity with skepticism is reasonable rather than unhinged.

Where the strong claim breaks down

Skepticism is warranted. The leap from fabrication is real and growing to therefore virality is largely fake and centrally coordinated is where the evidence stops and extrapolation takes over.

The documented cases describe a mess, not a machine. The synthetic content and inauthentic engagement that investigators actually find come from many separate actors with unrelated, mostly banal motives: marketing agencies building novelty personas, spam operators chasing ad payouts, engagement-for-hire services, and, in a minority of cases, state propaganda units. Each is explainable without a central conspirator, and they frequently work at cross purposes. No study, leak, or dataset has shown that the trends most people actually see are orchestrated by a single hand, which is what the strong version requires.

The theory also sits awkwardly with its own best evidence. The reason we know about large bot networks and covert AI operations is that they were caught and removed, and that the operations disclosed by AI companies were, by those companies' own accounts, often failing to build real audiences. A world in which platforms secretly manufacture all virality is hard to reconcile with a world in which platforms keep publishing takedown reports of the manufacturers. The detected cases are better read as a real, contained problem than as proof of a total, successful takeover.

And a great deal of virality remains demonstrably organic. Breakout songs, videos, memes, and accounts are routinely traced back to identifiable real people, sometimes uncomfortably so for the humans suddenly thrust into attention. The claim that “little of what trends is organic” is not a measured finding; it is an assumption applied in advance, and it is contradicted every time a viral moment turns out to belong to a real person who did not want it.

The unfalsifiable turn

It is worth isolating the specific move that turns a fair worry into an unprovable one, because it is the same move every time, and it is what separates the disputed claim from the documented one.

The genuine difficulty, that synthetic content is often hard to distinguish from human content, gets converted into a rule: if you cannot tell, assume it is fake. That rule feels prudent, but it makes the theory unfalsifiable. Real virality can be waved away as bots. Removed bot networks can be waved away as a distraction or a limited hangout. A synthetic persona that flops proves the danger; one that succeeds proves the takeover. A claim that absorbs every possible observation, in every direction, has stopped describing the world and started describing a mood.

The absence of visible fakery becomes, in this frame, evidence of sophisticated fakery, the same logic that lets any natural event be recast as a cover-up if one simply assumes a clever enough hidden actor. It is seductive precisely because it can never be wrong, and worthless for the same reason.

“You can't tell what's real anymore” is a true and serious statement. “Therefore none of it is real” is a different statement, and it does not follow.

The disciplined alternative is to hold the difficulty and the conclusion apart: to take detection, disclosure, and measurement seriously, to treat the share of synthetic content as an open empirical quantity rather than a foregone one, and to notice that the honest answer to “how much of virality is fake?” is “a real and growing amount, not yet cleanly measured, and not all of it.”

Why the sweeping version resonates

This is one of the rarer entries where the believers' starting instinct is empirically ahead of the dismissers'. Something did change, it is measurable, and people noticed it before they had a name for it. That solid foundation is exactly why the exaggerated version spreads so easily.

It rides a real before-and-after. Anyone who has used the same feeds for years can feel that they have grown more templated, more synthetic, more cluttered with content nobody seems to have made on purpose. That impression is accurate as far as it goes, which lends the totalizing claim borrowed credibility it has not separately earned.

It is fed by cheap tools in real time. Because generative AI became widely available just as the framing spread, believers encounter fresh apparent evidence daily: an uncanny video, a suspiciously fluent review, a persona that might be a person. Each new example reads as confirmation of a theory that predates the specific example, a uniquely persuasive pattern in which the prediction seems to keep coming true.

And it does useful emotional work. In a low-trust environment where opaque algorithms already decide what billions of people see, a single nameable cause (it is all fake and coordinated) resolves discomfort more neatly than the true, messier answer, that thousands of uncoordinated actors are each chasing narrow goals that add up to a stranger internet without anyone planning it. It also offers a standing excuse: anything popular that one dislikes can be dismissed as bots, which is a comfort the evidence does not require but the mind reaches for anyway.