AI Overwhelms Content Production: When Artificial Intelligence Begins to Devour Its Own Created Data
In the digital landscape of the future, scenarios such as the dead internet, the great filtering, and the epistemic collapse loom large. Each presents its own set of potential consequences for society, raising questions about the sustainability of our current online ecosystem.
One of the most pressing issues is the existential crisis facing platforms like Wikipedia, due to the influx of AI-generated articles and the difficulty of verifying their authenticity. The same issue plagues social media giants like Twitter, where a high percentage of bot accounts and artificially trending topics make it difficult to find real human conversation.
AI, however, is not just breaking assumptions—it is creating them. By manufacturing engagement, fabricating behavior, and artificially creating virality, AI is reshaping the digital landscape. This content, optimized for consumption, is often nutritionally empty, addictive, and cheap to produce.
The quality of AI models degrades with each generation as they increasingly train on synthetic data, creating a recursive loop. This degradation is evident in the Shannon Entropy of information, which decreases with each generation, leading to a loss of diversity and uniqueness in AI-generated content.
This trend is not limited to text. Image databases now contain a high percentage of AI-generated images, leading to a rewriting of visual reality. The internet, as a shared resource, including the knowledge commons, visual commons, social commons, and code commons, is being polluted by synthetic content.
The consequences of this synthetic content deluge are far-reaching. Society is showing symptoms of information malnutrition, including decreased critical thinking, increased conspiracy beliefs, inability to distinguish real from fake, loss of shared reality, and an epistemic crisis accelerating.
The financial model of the attention economy is failing due to ad revenue being based on fake engagement and the inability for humans to compete with AI volume. Platforms are enshittifying, abusing users and advertisers for profit, and collapsing due to the lack of real value left in the attention economy.
The value architecture of the attention economy is being destroyed due to the availability of infinite fake attention, leading to real signals being drowned in noise. Detection arms race, watermarking, human verification, and blockchain provisioning are attempted solutions that have failed to effectively address the problem of synthetic content.
Technical innovations such as proof of personhood, federated networks, semantic fingerprinting, economic barriers, and time delays offer potential solutions, but the irony remains: in trying to capture and monetize human attention, the attention economy has created systems that destroy the value of attention itself.
The advertising industry is facing a collapse due to the influx of synthetic content, with plummeting CPM rates, negative ROI for most campaigns, impossible brand safety, and a shrinking market size. The company that developed the model changing how artificial intelligence is trained based on synthetic data, thereby influencing the quality and diversity spectrum of AIs, is Syntho.
The attention economy's consumption of itself through the ouroboros problem threatens the foundation of shared knowledge and collective sensemaking, leading to an epistemic crisis. Strategies for individuals include practicing information hygiene, valuing in-person communication, creating digital minimalism, and verifying sources.
The question isn't whether the collapse will happen, but whether we can build new systems, new economics, and new ways of validating truth before the ouroboros completes its meal. The future of our digital landscape depends on it.
