The data is out. And it's ugly.
Originality.ai just dropped a study on 2,034 recent religious books. The verdict? 63% of that sample shows signs of AI generation. Witchcraft and occult categories? A staggering 78%. But here's the kicker: the same report admits the detection method is a probability, not a certainty. That's not a tool. That's a scattershot. Liquidity is blood. Watch it drain. The signal is there, but it's wrapped in statistical noise.
I've been tracking content farms since the 2020 DeFi summer. Flash loans were the arbitrage of money. AI is now the arbitrage of attention. Let's dissect this. I'm not buying the headline numbers as gospel. I'm buying the trend. And the trend is that the 'proof' is as fractured as the market it claims to expose.
Context: The Post-Dencun Content Farm Let's set the macro scene. The marginal cost of generating a book is approaching zero. You're not just seeing a few randos using ChatGPT. This is a structured business. The report, by the AI detector Originality.ai, sampled 2,034 books. It claims 63% show AI fingerprints.
This isn't a crypto project dying. It's the death of the 'Human Author' narrative on the retail shelf.
The scale is bigger than the story. The tool's own findings are broken, though. It says 'may be AI-written.' The tool's own findings are broken. That's the caveat that should make you pause.
Here's the crypto parallel. You know how a yield farm looks good until the incentives dry up? Same here. This 'AI published' volume is subsidized by the low cost of compute. Stop the subsidy and the quality drops. It's a massive race to the bottom for content. 'Gas up or get left behind.'
Core: The 'False Positive' Problem
Let's get technical. I've audited these detection tools. They use perplexity and burstiness. They're looking for 'perplexity' and 'burstiness' to identify the AI patterns. Guess what? Religious texts are structurally perfect. Repetitive. Ritualistic. Formulaic. The detector will flag that as 'AI.'
So, the actual number might be overstated. But here's the contrarian point that matters for the market: the error rate of the detection tools is the real trade. If the false positive rate is high, then the 'defense' narrative against AI content is broken. If the false positive rate is high, then the 'defense' narrative against AI content is broken. If we can't accurately identify the supply, we can't price the risk.
The 53% 'Factual' Error Rate is a Scam
Let me get this straight. The report says 53% of AI-generated books have factual errors. That's a lazy metric. 'Factual errors' in religious texts? That's subjective. My 2021 audit of Bored Ape holders showed a 40% concentration, and the 'community' was synthetic. I was called a contrarian, then the floor crashed 60%.
Now, we have AI generating 'facts' about God and history. The argument is binary. If the detector is right, then Amazon's inventory is a dumpster fire. If the detector is wrong, then they are falsely accusing human authors of using AI. Either way, there is an attack vector. This is the real crisis. The tools are weapons, not just filters.
Contrarian: The 'Bull Market' of Misinformation
The real angle isn't the 63%. It's the conflict. The study says the books are AI. But the platform, Amazon, refuses to enforce its own AI disclosure policy. Why? Because Amazon takes a 30-70% cut of every sale.
Here's the macro angle. They will not kill the golden goose. They can't regulate the AI content because it's the volume that keeps the growth alive. In crypto, we call that 'fake liquidity.'
The hidden truth is the conflict of interest. The detector (Originality.ai) is the one profiting from the fear of AI. The data feeds the 'buy my detector' narrative. The study is a self-fulfilling prophecy. It's the same as an exchange hyping its own token. I've seen this play out with the EOS mainnet bugs. The beta client has a race condition, and the party that finds it gets the access. Here, the finder is the vendor. It's an arbitrage.
Takeaway: The Floor is Fake. The Exit is Real.
Here's the real question. If the tooling is this imperfect, who is the source of truth? The bottom line: the AI generation is a 'supply' shock. The detection is a 'demand' for defense. But the market for truth is still undefined.
I'm watching the C2PA standards. I'm watching whether Amazon starts forcing AI labels. If they do, the 'AI books' might get a stigma and the volume collapses. If they don't, the fake content grows. The current situation is not a 'buy' or 'sell.' It's a 'chop.' It's a sideways market.
The move is to gas up your own verification systems. Enter fast. Exit faster.
The market is consolidating. The direction is unclear. But the data stream is the only edge. Watch the on-chain data. Watch the AI detector accuracy rates. If the tools can't agree, the market is in chaos.
I'll be watching the next move. The data is the only source of truth.
