The second quarter of 2025 will be remembered as the moment "AI disclosure" stopped being a theoretical policy debate and became an operational reality for every indie author on KDP. Three separate forces converged between April and June: Amazon's enforcement of its AI-content labeling rules reached critical mass, the first wave of publisher-to-AI-company licensing deals became public, and a federal court in California handed down the first substantive ruling on whether training on copyrighted books constitutes fair use. If Q1 was the year AI entered publishing's vocabulary, Q2 was when it entered the contracts.
Amazon enforces AI disclosure — and the gray zone gets uncomfortable
KDP's requirement to disclose "AI-generated" content had existed since late 2023, but enforcement was inconsistent and self-reported. By April 2025, Amazon began flagging titles retroactively and issuing warnings to publishers who had not disclosed. The policy distinguishes between AI-generated content (must be disclosed) and AI-assisted content (no disclosure required) — a line most authors found impossible to draw cleanly. The practical effect: authors using tools like Sudowrite, Claude, or ChatGPT for drafting are operating in a compliance gray zone that Amazon has not resolved. What this means for you: Document your workflow. If AI wrote sentences you kept, disclose. If AI suggested ideas you rewrote, you're in "assisted" territory — but keep records in case the policy tightens. Major publishers begin licensing backlists for AI training
Wiley, HarperCollins, and at least two mid-size academic publishers confirmed licensing agreements with AI companies for training data access during Q2. The deals were not announced publicly — they surfaced through industry reporting by The Bookseller and Publishers Marketplace. Authors in those catalogs were not individually notified. This set off a wave of contract scrutiny: rights advocates pointed out that most legacy publishing contracts signed before 2020 contain no explicit AI training clause, meaning publishers may be licensing rights they were never granted. What this means for you: If you have backlist titles under legacy contracts, request a rights audit. Organizations like the Authors Guild published template letters for requesting clarification from publishers. For new contracts: insist on an explicit AI training exclusion clause.
The Ingram vs. KDP print cost gap widens
A less-covered but practically significant development: IngramSpark adjusted its print pricing structure in May 2025, reducing per-unit costs for black-and-white trade paperbacks above 200 pages. Simultaneously, KDP Print held its pricing, widening the cost gap for authors running print editions at volume. For authors selling direct via Shopify or their own store, Ingram's unit economics became meaningfully better for orders above roughly 50 copies. What this means for you: If you're running any kind of direct sales operation and your print volume is growing, a side-by-side cost comparison is now worth doing. The calculation that favored KDP Print for simplicity in 2023 may no longer hold.
Looking ahead to Q3: The California fair-use ruling from June — which found in favor of the AI company on training data but against on output reproduction — was immediately appealed. Q3 will be defined by whether the appellate court fast-tracks the case, and by whether KDP's disclosure enforcement expands to audiobooks, where AI narration has proliferated with almost no oversight.