AI Will Make Pre-GPT Writing a Fetishized Authenticity Marker

Krzysztof Pelc argues that the arrival of generative large language models (LLMs) has created a permanent divide in all authored writing: texts produced before 2022 (the all-human era) and everything after. He predicts that readers will increasingly prize pre-GPT works for their guaranteed human origins, leading to a fetishization of the past. Publishers will mine back catalogs, reissue 'lost gems' with 'Verified Human Content' labels, and eventually forge new old works—abandoned manuscripts, recovered juvenilia—to satisfy the longing for authenticity. Pelc draws parallels to the Ossian forgeries (Napoleon's favorite poetry, actually invented by James Macpherson) and 18th-century sham ruins in English gardens. He notes that AI companies themselves are hoarding pre-LLM text: Google paid $60 million to Reddit, OpenAI licensed Time, Amazon licensed New York Times content. The real threat is not hallucination but the impossibility of knowing whether any contemporary text is human or synthetic. Recursion—training on AI output—flattens language and erases minority styles. Pelc suggests that doubt itself may be the last refuge of the human, as LLMs cannot genuinely express uncertainty.