The arrival of large language models has done something remarkable to book publishing: it tripled the number of new titles released monthly between 2022 and 2025, with some categories growing tenfold. But volume and value are proving to be very different things, and the gap between them contains lessons that extend far beyond publishing.
New research from Cornell and the University of Minnesota, drawing on over 333,000 Amazon ebook releases, delivers a nuanced verdict on AI’s creative impact. Average quality across all new releases has declined in the LLM era. The market has been flooded with low-quality content, primarily from new authors who entered the market precisely because AI lowered the barriers to doing so. These entrants produce predominantly poor-performing work, measured by reader engagement and ratings.
Yet the story contains a genuinely encouraging signal that most headlines will miss. The top 1,000 monthly releases per category have actually improved in quality compared to the pre-LLM period. Authors who were already established before LLMs arrived are producing higher-quality output than before. AI, in their hands, appears to function as a genuine creative complement rather than a substitute. The skill amplification hypothesis — that AI disproportionately benefits those who already possess domain expertise — finds real-world support here.
The mechanism matters enormously for how executives should interpret this pattern across industries. LLMs did not simply allow more “draws from the same urn.” They appear to have created two distinct populations: a large new cohort of low-capability entrants empowered to produce at all, and an existing cohort of skilled practitioners whose ceiling has risen. The average outcome deteriorates even as the frontier improves.
The welfare implications are more optimistic than the quality headlines suggest. The researchers’ market calibration estimates that AI-enhanced book production could raise consumer surplus from book markets by 25 to 50 percent in steady state, driven by a richer selection of genuinely valuable titles at the top of the distribution.
For investors and strategists, the broader implication is this: in any creative or knowledge-intensive domain disrupted by AI, average output quality may be a misleading indicator of AI’s true value creation. The action is at the tails. Organizations that can identify and amplify existing expertise will capture the upside. Those that simply lower barriers to entry will produce noise.
Source: Raw/trigger-ai-and-the-quantity-quality-of-creative-products.md