Synthetic Authority
How AI content farms steal a scientist’s credibility to sell you a political worldview with cumulative framing.
It began with a quiet Sunday morning, a coffee and a phone.
Neil deGrasse Tyson (I’m a big fan) had thoughts about the observable universe containing 10^80 atoms. Then I looked at who was actually posting: the account was called Xpiedon, a facts and science outlet that donates a portion of earnings to unspecified causes. Tyson’s photograph presided over a Facebook group titled Neil deGrasse Tyson, created in January 2024, 265,000 members, roughly one post every two hours, every day. What I found was not a fan community. It was an industrial operation, and a politically directed one.
Three feeder pages take turns filling the group: Xpiedon (Admin, 20,000 points), Factz Verse (Admin, 112,000 points), Science Tech Facts (Group Expert, 137,000 points). Together they maintain a near-continuous flow of content into an audience that believes, reasonably enough, that they are somewhere in the orbit of one of the world’s most recognisable scientists. None of it has anything to do with Neil deGrasse Tyson.
The prose has the texture of machine-written material: smooth, hedged just enough to avoid obvious error, plausible without being specific. No human team writes at this volume or this cost. Facebook’s Content Monetisation programme rewards it generously, the view counts translating into tens of thousands of dollars monthly, probably more, because the content has drifted toward categories that command premium rates: politics, health, social controversy, anything that makes people stop scrolling and type something angry. The product being sold is trust, borrowed from a man who has no idea it is happening.
The early posts are benign. Cosmology. Human origins. You build the room before you decide what to do with the people inside it. In a single morning’s scroll: Florida abolishing vaccine mandates, framed as a victory for personal liberty. A Trump administration response to a disease outbreak described as measured, freedom-preserving leadership. Three paragraphs platforming RFK Jr.’s chemtrail beliefs, followed by one paragraph of rebuttal that read like a legal disclaimer. Evolutionary science and biblical mythology presented as parallel research traditions. A photograph of Muslim women in full covering, overlaid with the word BANNED, and the caption: “Trump Proposes A Complete Ban On Muslims Entering The U.S. What Do You Think.” That post generated 142 comments.
Each post looks responsible. The cumulative effect is something quite different. Researchers call it cumulative framing: each post a small adjustment of a lens, and by the time you have read twenty of them across a week the lens has moved considerably without your noticing. You believe you have been following the news. What you have been receiving is a curated ideological programme, delivered in the language of scientific authority, by a machine, under a borrowed name.
This is not a fringe operation. In March 2026, NewsGuard identified over 3,000 AI content farm sites across sixteen languages, more than double the number from six months earlier. A Stanford study the previous year documented coordinated clusters of AI-generated pages achieving hundreds of millions of views on Facebook alone, with the platform’s own algorithm amplifying the content because engagement signals were high. Merriam-Webster named “AI slop” its Word of the Year in 2025. The phenomenon has a name because it is everywhere.
Consider whose name has been borrowed. Neil deGrasse Tyson is the executive producer of “Shot in the Arm,” a documentary scrutinising vaccine hesitancy, with RFK Jr. as a central case study. Tyson called the film “a vital rung in a ladder of science literacy that we all must ascend, lest civilisation teeter on the brink of collapse from its absence.” The group built on his photograph is platforming RFK Jr. sympathetically, multiple times a week, to 265,000 people who believe they are reading something connected to him. Brian Cox has faced the same thing, publicly condemning AI deepfakes putting conspiracy claims in his mouth and thanking YouTube in 2025 for removing them. The playbook is identical wherever it runs: borrow a scientist’s credibility, generate content at industrial scale, use the audience for purposes the scientist would find repugnant.
The operation is simple by nature. Take a trusted public identity, build a fan group around their name, populate it with harmless material until the audience is large, then introduce the payload. Facebook’s terms of service prohibit impersonation, but the group calls itself a fan community, links to Tyson’s real page, and ensures each individual element carries a defensible reading even as the totality does not. The enforcement model was built for a different problem. In the narrow legal sense, no fraud has been committed. In every meaningful sense it is precisely a con: a systematic operation extracting trust, political attention, and revenue, through deliberate deception about what it actually is.
265,000 people log in each morning to read a scientist who is not there. I’m no longer a fan of this fan page.
