Something strange happened to the internet in 2025, and most people didn't notice. For the first time, AI-generated content — text, images, video, code — exceeded the volume of human-created content published online. We crossed a threshold that researchers had been warning about for years: the internet is no longer primarily a record of human expression. It's increasingly a conversation between machines.
The Numbers Are Staggering
By some estimates, over 90% of all content on the internet will be AI-generated or AI-assisted by 2027. Already, vast portions of what appears in search results, social media feeds, and product reviews was written by language models. SEO content farms have replaced human writers almost entirely. E-commerce platforms are flooded with AI-generated product listings. Social media bots powered by GPT-class models are indistinguishable from real users.
Even platforms that try to stay human-first are losing the battle. Reddit implemented new bot detection systems in 2025 after researchers discovered that entire subreddit conversations were being conducted between AI accounts, generating engagement metrics that looked perfectly organic.
The Model Collapse Problem
Here's where it gets technically alarming. AI models are trained on internet data. If the internet is increasingly AI-generated, new models are being trained on the output of older models. Researchers call this model collapse — a degenerative process where each generation of models becomes slightly worse, like a photocopy of a photocopy.
A landmark 2023 paper demonstrated that models trained on model-generated data progressively lose the ability to represent the full diversity of human language. They converge toward a bland, generic mean. The tails of the distribution — unusual ideas, minority perspectives, creative expression — are the first to disappear.
The Trust Erosion
The practical consequences are already visible. Online reviews are increasingly useless — you can't tell which are genuine and which were generated by a seller's AI. News aggregation sites are filled with AI-rewritten versions of the same stories, sometimes with hallucinated details added. Academic research has found AI-generated papers slipping through peer review.
Trust in online information, already declining before generative AI, is in freefall. A 2025 Pew survey found that 63% of Americans "rarely or never" trust content they find online, up from 38% in 2022.
The Human Premium
A countertrend is emerging: verified human content is becoming valuable precisely because it's rare. Platforms like Substack, where individual writers build audiences based on personal voice and accountability, are thriving. Podcasts — hard to fake convincingly at scale — are gaining trust as a news source. Some publications have started advertising "100% human-written" as a selling point.
The irony is sharp. We built AI to generate content at scale, and in doing so, we may have made the most valuable content the kind that AI can't convincingly produce: work with genuine voice, lived experience, and human accountability.
What Comes After the Synthetic Flood
The optimistic view is that the internet will bifurcate into a "synthetic web" of commoditized AI content and a "human web" of authenticated, valued personal expression. The pessimistic view is that the noise will drown out the signal entirely, and we'll lose the most remarkable information-sharing system humanity has ever built.
The realistic view is probably somewhere in between — but it requires action. Content authentication, platform accountability, and a cultural shift toward valuing provenance over volume. The alternative is an internet that's technically bigger than ever but practically useless.