AI-generated content is suffocating independent digital publishing through brute volume rather than literary merit. An empirical study of 14,419 self-published Amazon e-books released between January 2023 and March 2026 shows that automated mass production is actively wrecking unit economics across the digital shelves. To map the fallout, researchers benchmarked titles against daily sales data from a Big Five US publisher tracking roughly 500,000 Amazon titles, capturing approximately 95 percent of daily e-book revenue.
Rather than skimming preview snippets, the researchers evaluated full texts using the Pangram v3.3 detector, which operates at a documented 0.04 percent false-positive rate. Categorizing catalog inventory by machine-generated share—none, light (up to 25 percent), and substantial (over 25 percent)—reveals that heavily automated titles captured 20 percent of total volume, 12.1 percent of unit sales, and 11.3 percent of gross revenue. Pure human titles (62.9 percent of catalog volume) still claimed 72.5 percent of gross revenue, confirming that consumers prefer human writing. Yet sheer volume broke the distribution engine.
Between Q1 2023 and Q1 2026, the cumulative catalog grew 38.3x while the number of titles selling per quarter grew 19.2x.
Quarterly revenue grew by just 8.9x over the same window. When catalog supply expands more than four times faster than buyer spending, discovery algorithms inevitably collapse under the weight of synthetic filler, dragging downstream discoverability with them.
Dilution Across Genre Categories
This supply shock directly penalizes human writers who never touched an AI generator. Comparing 2023 and 2025 release cohorts over identical post-launch windows, revenue per book declined across six of eight genres. When isolating titles with zero detected AI text, revenue per title still collapsed in seven of eight genres. Passive marketplace moderation has turned open content pipelines into a classic lemon market, where search dilution hurts every participant regardless of catalog quality.
The sole segment bucking the trend was Fantasy/Supernatural/Horror, where automated text entered latest and penetration remained lowest. There, revenue per title for non-AI books rose 35 percent, demonstrating that the downturn stems from synthetic supply saturation rather than broad macroeconomic weakness. The damage proved sharpest in high-density Kindle Unlimited categories where payouts draw from a shared subscription pool: human authors saw their revenue-share advantage compress by 8.4 percentage points compared to low-subscription niches. For open content platforms, passive ingestion is no longer viable; without rigorous provenance tracking and aggressive filtering at intake, automated zero-marginal-cost generation will steadily degrade the marketplace ecosystem.