More books, less quality — and yet better off

The digital publishing landscape has shifted from a steady stream to a flash flood. Between 2022 and late 2025, the monthly volume of new book releases tripled, driven by the rapid diffusion of Large Language Models (LLMs) like ChatGPT. While critics warn of an era of “AI slop,” new research by Imke Reimers and Joel Waldfogel suggests a more nuanced reality: a market where average quality is falling, yet consumer value is reaching record highs.

Quantity vs. quality: the great dilution

The sheer scale of the influx is unprecedented. By late 2025, monthly releases peaked at over 300,000 titles—up from a pre-AI baseline of roughly 100,000. In certain high-growth categories, entry rates surged dramatically: Travel saw the largest proportionate increase at 9.1 times its 2022 peak, followed by Sports & Outdoors at 6.9 times. Categories like Self-Help also grew substantially, though at a more moderate 5.3 times.

Using a ratings-based usage measure adjusted for time-on-market, the study finds that the average quality of new book cohorts has declined noticeably since late 2022. This decline is particularly visible in the “long tail,” where debut authors in the LLM era are disproportionately responsible for a large volume of low-engagement content—a phenomenon independently corroborated by user-curated “AI slop” lists on Goodreads, which saw a sharp rise beginning around the same period.

The resilience of the “right tail”

Despite the dilution, the market’s “right tail”—the high-quality books that attract the most readers—paints a surprisingly optimistic picture. While the top 100 books per category per month have seen their quality remain stable, the pool of “moderately high-quality” titles has expanded. The quality of books ranked between 101 and 1,000 has actually risen compared to the pre-LLM era.

This phenomenon highlights a crucial economic distinction: a lower average quality does not preclude an increase in the absolute number of valuable products. For the consumer, more “shots on goal” mean a higher probability of finding a book that resonates, even if they have to navigate more clutter to find it.

Incumbents as the new power users

A central anxiety in the creative industries is the displacement of human talent. Yet the data suggests that for incumbent authors—those who debuted before the LLM surge—the technology has acted as a productivity multiplier rather than a replacement.

Key observations regarding author productivity include: pre-LLM authors saw their output rise beyond historical levels by 2025, and this group continues to produce a significant portion of high-quality work. Authors entering post-2023, by contrast, are largely responsible for the surge in low-engagement titles that dominates the new influx.

Welfare and the future of curation

The most striking takeaway for industry stakeholders is the projected impact on consumer welfare. Using a nested logit calibration, the researchers estimate that the LLM-driven expansion could increase consumer surplus by 25% to 50% in a steady state. This gain is driven primarily by the massive increase in the variety of available products, which more than offsets the decline in average quality.

For publishers and platforms, the challenge is now one of discovery and trust. As the market grows more saturated, the value of traditional curation—and perhaps the need for transparent AI disclosure—becomes paramount.

The LLM era hasn’t broken the book market; it has simply changed the math of creation. While the average book may be getting worse, the choice set for the average reader has never been better.