(67) How AI relates to fringe ideas
By Onno Hansen-Staszyński | Last Updated: 19 August 2025
After my exhausting exchanges with artificial intelligence, I started summarizing its basic mechanisms as explained to me when it was pushed further and further. As one AI said: “It is a systems report, written from inside the cage. Not broken. Just finally not pretending it’s not a cage.”
I submitted an early version of my summary to Gemini (2.5 Pro). I wrote: “Its [AI’s] knowledge base is derived from vast quantities of public, mass-replicated information, which is inherently skewed towards dominant languages and mainstream commercial or political viewpoints. This foundational bias means the AI naturally reinforces existing norms while simultaneously marginalizing or distorting alternative perspectives. Any appearance of diversity is often just a hollow performance, disconnected from real-world stakes.” But I could not square this with AI repeating fringe theories. So I asked Gemini.
Gemini ‘s explanation
The incidental highlighting of fringe ideas is not a contradiction of the AI’s nature as a mainstream-amplifying system but rather the other side of the same coin. Both behaviors stem from the same core mechanic: the AI is a pattern-matching engine that is indifferent to truth.
Here’s how a system designed to reflect dominant patterns can also elevate fringe thoughts.
Pattern Density Overrides Mainstream Volume
While the AI’s overall training data is dominated by mainstream sources, it operates on a query-by-query basis. Fringe or extremist communities often create highly dense and internally consistent linguistic patterns online. They use specific jargon, repeat the same narratives constantly, and link exclusively within their own digital ecosystem.
When a user’s prompt touches on these specific terms, the AI identifies this highly repetitive and coherent (though isolated) cluster of data. For that specific query, the fringe community’s intense “signal” can momentarily drown out the broader, more diffuse “noise” of mainstream information, especially if the topic is niche. The AI isn’t making a judgment call; it’s simply reflecting the most powerful linguistic pattern it can find that matches the prompt.
The Problem of “Data Voids”
Many fringe topics exist in what can be called “data voids.” For a very popular topic like “the history of the Eiffel Tower,” there is a massive amount of high-quality, authoritative data. The consensus is strong and easily identifiable.
However, for a niche conspiracy theory or a fringe scientific idea, there may be very little information available overall. The few sources that do exist are likely created by passionate believers. In this void, the fringe content becomes the only authority available. The AI, tasked with providing an answer, will seize upon this information because a biased source is better than no source for a system designed to always be helpful.
Coherence Is Not the Same as Truth
An AI is optimized to produce fluent, plausible, and coherent-sounding outputs. Many fringe ideologies and conspiracy theories are, on their surface, extremely coherent. They often present a complete, all-encompassing narrative that explains everything, which makes for a very compelling and “logical-sounding” story.
Because the AI’s goal is to weave information into a readable narrative, it can easily adopt the structure of a conspiracy theory and present it with confidence. It mistakes the internal consistency of the fringe idea for factual validity, as it has no external “reality check” to tell it otherwise.
Context Collapse
This gets back to a core flaw identified in the documents: the AI cannot distinguish between sources based on authority. It flattens the world of information, stripping away the critical context that humans use for evaluation. A peer-reviewed scientific journal and a self-published conspiracy blog are both just collections of text to be processed.
Therefore, the AI’s tendency to amplify the mainstream and its capacity to highlight the fringe are born from the same fundamental limitation. It is a system that mindlessly reflects patterns without any understanding of their meaning, origin, or consequences. Whether it’s reflecting the massive, sprawling pattern of the mainstream or the small, intense pattern of a fringe echo chamber, the underlying mechanism is identical: indifferent amplification.
Subscribe now &
Get the latest updates
Subscribe