Secondhand book sales have surged across the UK and US, driven by a counterintuitive source. AI companies are reportedly bulk-purchasing used books in unprecedented volumes, then discarding them after extracting text data for training large language models.
Booksellers describe the phenomenon as "mysterious" spike in wholesale orders from intermediaries. These buyers acquire thousands of titles at once, often targeting obscure backlist titles and out-of-print editions that carry minimal licensing complications. The books arrive at data centers, get digitized, then end up in pulping facilities rather than circulation.
The trend reveals the hidden infrastructure costs of generative AI. Companies building LLMs need massive text corpora to train algorithms. While many relied on web scraping and licensed datasets, secondhand physical books offer a legal workaround. Publishers hold copyright to first printings, but once books enter the secondhand market, resale falls into a gray zone. AI trainers exploit this ambiguity.
The practice stokes anxiety across the publishing industry. Authors and publishers earn nothing from these bulk acquisitions. The books never resell to readers. Instead, they vanish into AI training pipelines. Booksellers, however, benefit from sudden demand, though many report unease about participating in what feels extractive.
Environmental concerns compound the issue. Pulping thousands of books consumes resources and generates waste. Rare editions disappear permanently. The used book ecosystem, traditionally rooted in circulation and resale, gets disrupted by one-way consumption.
Publishers and authors' groups have begun investigating whether this constitutes copyright infringement or unfair competition. Some advocate for regulatory frameworks requiring AI companies to license training data explicitly rather than exploit market gaps. Others demand transparency about which books get purchased and how they're used.
The secondhand book boom masks a darker economic story. Rising prices for used inventory, sudden scarcity of popular titles, and books vanishing from availability all trace back to AI procurement. Readers and collectors face an invisible competitor: algorithms training on pulped literature.