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

**No verdict.** The evidenced core is substantially real: AI-generated influencers with genuine audiences exist, mass-produced synthetic content (widely called 'AI slop') really is flooding platforms, and bot networks and paid engagement farms that fabricate likes, views, and followers are well documented, with platforms removing them by the hundreds of millions. The claim we rate is the stronger one: that essentially all virality is now centrally fabricated and that organic human culture online is effectively gone or controlled. That totalizing version is unproven and, as usually stated, unfalsifiable, and plenty of virality remains demonstrably organic. A real, growing, evidenced phenomenon has been exaggerated into a sweeping claim the evidence does not support.

Category: Science, Space & Technology · Era: 2020s · First circulated: Consolidated across 2024-2025 as 'AI slop,' AI influencers, and inauthentic-engagement stories converged; the framing that little of what trends is organic sharpened in 2025-2026 · Believed by: A fast-growing share of heavy social-media users, especially younger ones who encounter synthetic content daily; it overlaps with, but is more specific and more current than, the broader Dead Internet Theory
URL: https://theconspiratory.com/theory/ai-fake-virality

## Summary
The claim that online virality has quietly become fake: that AI-generated influencers, synthetic video and voice personas, bot armies, and paid engagement farms now secretly manufacture the trends, 'viral' moments, and popular accounts people see, so that little of what trends is organic. Much of the documented core is true. Openly AI-generated influencers earn real money and real followings, mass-produced 'AI slop' is measurably flooding feeds, and platforms have removed enormous numbers of fake accounts and engagement. What overreaches is the strong version, that essentially all of it is centrally coordinated and that organic human culture online is gone. This file separates the evidenced trend from the sweeping conspiracy. It is a close cousin of the older, broader Dead Internet Theory, narrowed to the specific 2025-2026 claim about manufactured virality.

## The claim
That the viral trends, breakout accounts, and 'popular' content people encounter online are now largely manufactured rather than organic: produced or amplified by AI-generated influencers, synthetic video and voice personas, automated bot networks, and paid engagement farms, and in the strongest version coordinated centrally so that little of what appears to be spontaneous human culture actually is.

## Origin and timeline
- 2021: The Dead Internet Theory is named in a forum post arguing the web is mostly bots and AI. It supplies the broad template, that human activity online has been displaced by machines, that the later, more specific 'fake virality' claim narrows and updates.
- 2022–2023: Consumer generative-AI tools for text, image, voice, and eventually video become cheap and widely available, making convincing synthetic personas and content producible at scale by almost anyone.
- 2023-11: The Spanish agency The Clueless publicizes Aitana Lopez, a fully AI-generated 'model' presented as a 26-year-old from Barcelona, reported to earn up to 10,000 euros a month from brand deals. Coverage frames her as proof that an entirely synthetic influencer can draw a real audience.
- 2024: 404 Media documents how Facebook 'AI slop' pages are produced, reporting that creators in several countries follow guides and chase platform performance bonuses to farm engagement with mass-generated images. The term 'AI slop' spreads into common use.
- 2024-05: OpenAI discloses that it disrupted five covert influence operations, linked to Russia, China, Iran, and a private firm, that had used its models to generate comments, fake biographies, and propaganda: a documented case of AI being used to fabricate the appearance of grassroots opinion.
- 2025-01: AI-generated 'user' profiles surface on Facebook and Instagram after a Meta executive describes wanting AI accounts to exist alongside human ones. A public backlash follows and the profiles are removed within days, crystallizing fears that platforms themselves will populate feeds with synthetic personas.
- 2025: AI slop goes mainstream. By some estimates roughly half of new English-language web articles are AI-generated; Spotify reports removing tens of millions of spam tracks; a notable share of the fastest-growing YouTube channels are AI-made. 'AI slop' is named a word of the year.
- 2025-09: An AI 'actress,' Tilly Norwood, generates industry backlash after her creators seek representation, sharpening the sense that synthetic personas are being pushed into culture whether audiences want them or not.
- 2025–2026: The strands consolidate into a single popular claim: that virality itself is now largely fake and manufactured. The framing spreads faster than careful measurement of how much of what people actually see is synthetic.

## The evidence, claim by claim
- Claim: Fully AI-generated influencers with real audiences and real income already exist.
  Evidence: Documented, and openly so. Agencies have built synthetic 'models' such as Aitana Lopez, presented as AI from the start, that have drawn hundreds of thousands of followers and reported monthly earnings from brand deals. Separately, an AI 'actress' project drew wide coverage in 2025. These are not hidden: they are marketed as artificial. They establish that a non-human persona can accumulate genuine engagement, which is the real kernel under the theory.
- Claim: Platforms are being flooded with mass-produced synthetic content, sometimes financially incentivized.
  Evidence: Well documented. Reporting has traced how 'AI slop' pages are produced at scale and how platform payout structures reward the engagement they farm. Independent estimates in 2025 put AI-generated material at roughly half of new English-language web articles, and streaming and video platforms have reported removing tens of millions of spam or synthetic uploads. The volume of synthetic content is real and rising.
- Claim: Bot networks and paid engagement farms manufacture likes, views, and followers, and coordinated AI operations exist.
  Evidence: Confirmed by platform enforcement and company disclosures. Services selling fake engagement are long documented, platforms report removing hundreds of millions of fake accounts, and AI companies have disclosed disrupting covert, state-linked operations that used their models to fabricate grassroots-looking opinion. Inauthentic amplification is a real, evidenced problem, not a rumor.
- Claim: Therefore essentially all virality is now centrally fabricated, and organic human culture online is effectively gone.
  Evidence: Not supported. The documented cases describe a large, messy mix of separate actors (marketing agencies, spam operators, engagement-for-hire services, state propaganda units) each with its own mundane motive, not one directing hand fabricating culture wholesale. No study, leak, or dataset has shown that most virality is centrally coordinated, and platforms keep removing the operations that are caught, which sits awkwardly with the claim that they secretly run everything. Plenty of viral moments remain traceable to identifiable real people.
- Claim: You can no longer tell real from synthetic, so it is safest to assume it is all fake.
  Evidence: The difficulty is genuine; the conclusion overreaches. That synthetic content is often hard to spot is true and worth worrying about. But 'hard to distinguish' is not 'therefore all fake,' and treating unfalsifiability as proof runs the wrong way: a claim that can absorb any evidence (real virality is dismissed as bots, removed bot networks are dismissed as a distraction) has stopped being testable. The honest position tracks how much is synthetic rather than assuming the maximum.

## Why people believe it
- The documented cases are real and easy to point to. Openly artificial influencers, financially incentivized 'AI slop,' and disclosed bot takedowns give the theory concrete, verifiable exhibits, which most conspiracy theories lack, so the leap to 'then most of it must be fake' feels small.
- Almost anyone who has used the same platforms for years has a genuine before-and-after: feeds that once felt handmade now feel templated and synthetic, and that lived impression reads as confirmation of the sweeping version.
- Cheap generative AI arrived just as the framing spread, so believers watch fresh 'evidence' (AI reviews, synthetic personas, uncanny viral videos) seem to compound in real time, which is unusually persuasive.
- Opaque recommendation systems already decide what billions of people see, and few outsiders can audit them, so a hidden-hand explanation feels simpler than a hundred uncoordinated profit motives.
- The strong version is also emotionally useful: any viral thing a person dislikes or distrusts can be waved away as bots or 'slop,' which is comforting even when it is not true, and comfort sustains belief.

## Open questions
- How much of what an actual person encounters in a feed, a trending list, or a comment section (as opposed to raw content volume) is synthetic or inauthentically amplified is still not cleanly measured. The direction is upward, but no independent study has published a reliable 'share of what humans actually see' figure, so the gap between measured slop and lived experience is genuinely unresolved.
- Whether reliable, at-scale detection of AI-generated text, images, and video is achievable, or whether generation and detection stay locked in a permanent arms race with detection a step behind, is contested among researchers. Until it settles, the question people now ask of any viral post (is this real?) may lack a stable answer for years.
- Whether autonomous AI influencers will gain genuinely large, durable audiences, or whether current examples are a marketed novelty that plateaus, is unsettled. Their reported reach is real but modest next to top human creators, and how far it scales is unknown.
- How much coordinated manufacturing actually succeeds, versus gets caught and removed or simply fails to catch on, is unclear. The best-documented covert operations are ones that were detected and, by their disruptors' own accounts, often failed to build real audiences, which may reflect either the true ceiling of such efforts or only the limits of detection so far.

## Latest developments
- 2026-08-18T09:40Z: A specimen that supplies the part this file's thesis usually has to argue for: the motive. Boatos rated as a hoax on 17 August a set of videos circulating in Brazil showing a woman in a uniform resembling that of Atacadao, a large wholesale supermarket chain, presented as a shelf-stacker who had become known online for adult content filmed in store, complete with simulated interview dialogue about customers recognising her. Boatos found no evidence that the woman is a real employee of the chain or that anything was recorded there, and identifies the videos as AI-generated. Two things are worth pulling out. Some versions direct viewers onward to other profiles and pages promising more of the supposed employee, which is the funnel, and a manufactured persona with a funnel attached is not a prank but a business. And the target has moved down-market. The synthetic personas this file was built on impersonated influencers and public figures; this one invents an ordinary worker at a named real company, which is both cheaper to fabricate and less likely to be checked, because nobody has a prior mental image of her to contradict. No real person has been shown to be involved, and this entry names none. (source: https://www.boatos.org/brasil/repositora-do-atacadao-viraliza-apos-seus-conteudos-gravados-no-local.html)
- 2026-08-09T15:30Z: A case that corrects something in the two entries below. Snopes rated fake on 9 August a ten-second clip shared as footage of a zombie attacking partygoers in an Atlanta nightclub, and got an unusually direct answer: a person managing the Facebook page Go Viral AI told Snopes by message that they had made it with OpenAI's Sora. But the important detail is the chronology. The original post, on 10 July, carried the caption 'Imagine partying and having fun then you see this....' and had its location set to Los Angeles. It was other users, reposting three days later, who attached 'Just Happened In a Atlanta Night Club' and named a venue. So the video and the false claim had different authors: the maker published a hypothetical and someone downstream converted it into a news report. That reframes the disclosure question this file has been tracking. It is not only that makers label their work inconsistently and that labels get stripped in transit; the harmful assertion can be written by a person who did not make the footage and has no obligation to anything the maker said. The named venue is checkable and wrong, which is the other lesson: a club of that name in Atlanta closed in 2007 after fatal shootings, per the Atlanta Journal-Constitution, and has not reopened. Detection was mundane, the growling far too clear for the distance from the phone and limbs contorting as people flee, and Facebook's own page-transparency data lists the account as managed from Indonesia, posting a stream of eight, ten and fifteen-second clips, all durations typical of generated video. (source: https://www.snopes.com/fact-check/zombie-attack-atlanta-nightclub/)
- 2026-08-09T03:20Z: A companion case to the entry below, and it cuts the other way on the question of disclaimers. Snopes rated Fake on 8 August an image presented as surveillance stills of a gorilla screaming for help after his favourite caretaker collapsed during a belly rub. OpenAI's detection tool, which reads the invisible SynthID watermark, identified it as generated, and Snopes noted the ordinary tells alongside that: distortion near the man's armpit and the gorilla's foot, and a surveillance timestamp on an image far sharper than surveillance footage ever is. The story named a caretaker and left both the gorilla and the sanctuary unnamed, which is what a fabricated account looks like when nobody expects it to be checked. Three things are worth carrying forward. The image came from a Facebook account, StoryTime, whose AI-generated animal stories Snopes has debunked repeatedly, including a December 2025 claim that a mother bear and cub rescued a lost girl in Alaska: this is a production line with a back catalogue, not a one-off. Snopes contacted the account to ask why it had published without any disclaimer, which sits against the two cases logged in the last 48 hours where the maker did label the work and the label was stripped in transit. Maker-side disclosure, in other words, is not just fragile once content is shared; it is optional and applied inconsistently by the same class of account. And Snopes reached for an older word for the genre, glurge, the fabricated sentimental story passed around as true. The format predates generative tools by decades. What has changed is that it now arrives with pictures. (source: https://www.snopes.com/fact-check/gorilla-caretaker-screaming-help/)
- 2026-08-08T15:20Z: A real event that produces no pictures turns out to be the ideal habitat for this, and 5 August supplied one. The upper stage of a SpaceX rocket that had been orbiting the Moon since January 2025 struck the lunar surface that day, as scientists had predicted in advance. The European Southern Observatory reported that its Very Large Telescope detected the chemical fingerprints of sodium and lithium around the time of impact, consistent with a plume lasting five to ten minutes. NASA had said beforehand that it would attempt to observe the impact from the ground while cautioning that weather and lighting might make viewing difficult. The confirmation, in other words, was spectroscopic. There was no footage, and there was never going to be. Within the hour a video of an object hitting the Moon and throwing up a plume was travelling widely on X under captions such as 'A SpaceX rocket has crashed into the Moon, leaving a 100 foot crater', and Snopes rated it fake on 8 August. The part that belongs in this file is the supply chain rather than the clip. A reverse image search traced the video to a version posted in April 2026, four months before the impact, by a Facebook page called Moon Watchers Club whose own bio reads 'Inspired by real events' and which puts out a steady stream of similar AI clips of objects striking the Moon, most of them washed in the same blue tint. The TikTok upload that first billed it as the SpaceX crash is timestamped less than fifteen minutes after the expected impact time. Nobody generated anything on the day. The footage was already in inventory, waiting for an occasion to be attached to, which is a more industrial thing than a fake made in response to the news. (source: https://www.snopes.com/fact-check/spacex-rocket-crash-moon-video/)
- 2026-08-04T21:10Z: A case checked on 4 August shows the synthetic-content problem being used as a vehicle for a straightforward racial and religious smear, which is worth logging because the technique and the payload are separable. A video posted to X on 3 August appeared to show a group of Muslim men kneeling on prayer rugs aboard a London Underground train, captioned as Muslims blocking the Tube and frustrating commuters, with a reference to Sharia law and the London mayor. Lead Stories found it AI-generated: the Hive Moderation detector returned 99.9 percent, and the clip carries the usual tells, a hand with six fingers, a finger bent in an anatomically impossible way, mismatched shoes, shoes laced from the toe upward, a carriage interior that matches no train in service, and prayer-rug patterns unlike real ones. Nothing depicted happened. This file rates how manufactured virality works rather than any single clip, and the mechanism here is the ordinary one: an emotive scene that confirms a prior costs nothing to generate and travels before anyone counts fingers. That it was built to carry an anti-Muslim claim is the part that does the damage after the fake is exposed. (source: https://leadstories.com/hoax-alert/2026/08/fact-check-fake-video-shows-muslim-men-praying-on-the-london-tube.html)
- 2026-07-18T21:02Z: A parallel wave hit health content. AI deepfake videos cloning the faces and voices of real, named doctors spread across TikTok and other platforms to push supplements and false medical advice, often aimed at women going through menopause. Fact-checkers and STAT documented hundreds of such clips, in which authentic footage of well-known physicians was altered so the fabricated version endorsed products the doctors never backed. It is the medical-authority version of the same synthetic-video problem: a trusted face cloned to sell an unverified claim. (source: https://www.statnews.com/2026/07/14/deepfakes-ai-generated-medicine-misinformation-regulation-video/)
- 2026-07-18T16:00Z: By mid-July 2026 the pattern had inverted. When Senator Mitch McConnell's office released a genuine photo to counter rumors about his health, users branded the real image an AI fake, and X's Grok chatbot amplified the claim while citing a newspaper 'debunk' that did not exist. Google's SynthID watermark and PolitiFact's forensic review confirmed the photo was authentic. It is the liar's dividend in reverse: a real image discredited by a fabricated AI accusation, with a chatbot inventing the evidence. (source: https://www.politifact.com/factchecks/2026/jul/13/tweets/mitch-mcconnell-rehabilitation-photo-ai-reused/)
- 2026-07-18T15:00Z: In July 2026 a wave of AI-generated videos spread across the UK, staging fake 'patriotic' protests and a migrant supposedly vowing 'the end of England,' pushing a false narrative of national collapse. Fact-checkers including Full Fact and Lead Stories identified the clips as AI fabrications, several carrying SynthID watermarks and visual glitches. The videos are fake, and the great-replacement 'takeover' story they were built to sell is a xenophobic propaganda trope, not documented reality. (source: https://fullfact.org/culture-and-society/muslims-march-halal-standards-fake-AI/)

## Sources
- Videos of a supposed Atacadao shelf-stacker are AI-created fakes, Boatos.org (2026): https://www.boatos.org/brasil/repositora-do-atacadao-viraliza-apos-seus-conteudos-gravados-no-local.html
- Video allegedly showing zombie attack in Atlanta nightclub is fake (the maker's Sora admission, the 'Imagine...' original caption, and the closed venue named downstream), Snopes (2026): https://www.snopes.com/fact-check/zombie-attack-atlanta-nightclub/
- Watch out for fake image of gorilla 'screaming for help' after caretaker's collapse (SynthID detection, the StoryTime account's back catalogue, and glurge as the genre), Snopes (2026): https://www.snopes.com/fact-check/gorilla-caretaker-screaming-help/
- Where Facebook's AI Slop Comes From, 404 Media (Jason Koebler) (2024): https://www.404media.co/where-facebooks-ai-slop-comes-from/
- Meet the Spanish AI model earning up to 10,000 euros a month, Euronews (2024): https://www.euronews.com/next/2024/12/27/meet-the-first-spanish-ai-model-earning-up-to-10000-per-month
- Facebook and Instagram to Unleash AI-Generated 'Users' No One Asked For, Rolling Stone (2025): https://www.rollingstone.com/culture/culture-news/meta-ai-users-facebook-instagram-1235221430/
- AI Actress Tilly Norwood Has Hollywood Fuming. Is She a Threat or a Stunt?, Variety (2025): https://variety.com/2025/film/news/tilly-norwood-ai-eline-van-der-velden-1236543362/
- 2025 was the year AI slop went mainstream. Is the internet ready to grow up now?, Euronews (2025): https://www.euronews.com/next/2025/12/28/2025-was-the-year-ai-slop-went-mainstream-is-the-internet-ready-to-grow-up-now
- Some of the largest online propaganda campaigns are using 'AI slop,' researchers say, NBC News (2025): https://www.nbcnews.com/tech/security/online-propaganda-campaigns-are-using-ai-slop-researchers-say-rcna244618
- Disrupting deceptive uses of AI by covert influence operations, OpenAI (2024): https://openai.com/index/disrupting-deceptive-uses-of-AI-by-covert-influence-operations/
- Artificial influencers and the dead internet theory, AI & Society, vol. 40 (Yoshija Walter) (2024): https://link.springer.com/article/10.1007/s00146-023-01857-0

Rated by The Conspiratory, a neutral, sourced encyclopedia of conspiracy theories. Full page: https://theconspiratory.com/theory/ai-fake-virality