The chart that’s been circulating among self-published authors this year comes from a working paper, not a press release, and it shows something specific: how many new ebooks Amazon added to its store each month, tracked from before generative AI tools were widely available through late 2025. The line does not creep upward. It roughly triples.
What the data actually shows
Economists Imke Reimers and Joel Waldfogel, in a National Bureau of Economic Research working paper first circulated in January 2026 and revised in May, used Amazon release data to track monthly new ebook volume. Their finding, summarized in the NBER Digest: monthly new ebook releases averaged around 100,000 in the 2020-2022 period and rose to more than 300,000 a month by late 2025. That’s a monthly rate that roughly tripled over several years, not a single before-and-after cliff the day ChatGPT launched, but the timing and shape of the increase track closely enough with the availability of consumer AI writing tools that the researchers treat generative AI as the primary driver.
What tripling actually means for a storefront
Three hundred thousand new titles a month is a substantial volume for a retailer to absorb. It changes the scale of the catalogue readers must navigate, alongside the task of enforcing content policies. The growth includes human-authored books as well as work produced with AI. The release count alone cannot tell a reader which process produced a particular title, or how well any review system handled it.
What Amazon actually changed
One response, reported by Publishers Weekly, was a cap introduced in September 2023 limiting authors to three new title uploads per day through Kindle Direct Publishing. That was an early measure; caps can change. KDP’s published book-creation guidance now specifies two new titles per book format each week. Such limits manage submission volume. They do not determine a manuscript’s authorship, and they operate alongside Amazon’s content policies rather than replacing them.
The identity check that exists, and what it doesn’t check
Separately, KDP’s identity-verification documentation describes checks for some accounts. Establishing who controls an account serves a different purpose from assessing a book. A verified identity does not show who wrote every passage, and an upload cap does not establish quality. Those limits are reasons to examine the separate content-review process, rather than evidence that Amazon never examines content.
The content review Amazon also describes
KDP’s Content Guidelines say enforcement uses machine learning, automation and human reviewers, and that reported concerns are investigated. They require publishers to disclose AI-generated text, images or translations to KDP, while distinguishing those from AI-assisted work. Content can be rejected or removed for failing the guidelines. These are content checks, even though the policy does not demonstrate that every violation is detected or every manuscript receives the same depth of review.
What the numbers actually settle
The tripling in monthly ebook volume does not prove that a specific book was AI-written. Nor does it measure Amazon’s screening success. What it establishes is a larger flow of releases, which gives a retailer more content to manage and readers more titles to compare. Evaluating the response requires separate evidence about submission controls, authorship disclosures and the effectiveness of policy enforcement.
What readers are actually left with
For a reader browsing the storefront, those distinctions still leave a practical question: what evidence supports trusting this particular book? An identity check offers information about the account holder; content-policy enforcement addresses compliance; neither amounts to a public certificate that every sentence was written by that person. Reviews, sample chapters and an author’s reputation can help a reader assess a title, while remaining imperfect indicators themselves.
The legitimate concern is the reach and effectiveness of existing review. A policy can set expectations and describe enforcement without publishing a comprehensive account of what it catches, what it misses or how often mistakes are reversed. Evaluating the process would require information about those outcomes as well as the number of books moving through it. That is the evidence a reader or author would need to assess whether enforcement is keeping pace with the expanding catalogue.
The available release data and policy documents do not quantify the proportion of books meaningfully assessed for authorship. Without that measure, the size of an enforcement gap cannot be calculated simply by comparing monthly releases with the existence of a daily upload cap. The volume growth makes the question more pressing; it does not supply the missing performance data.
Better tools, clearer disclosures and stronger enforcement could all affect that performance. Each would need to be evaluated against its intended purpose. A mechanism that reduces bulk submissions might help a review team manage workload without identifying AI-generated prose. A mechanism that examines content might find a policy violation without resolving who drafted it. Treating those outcomes separately makes it possible to ask whether a particular intervention works.
For independent authors, this is also a discoverability problem. More releases can mean more competition for a reader’s attention even when all the books being compared comply with the rules. Quality enforcement and discovery therefore overlap without being identical: removing misleading material does not automatically make a carefully written book easy to find. The catalogue’s expansion can matter commercially even when there is no policy violation to investigate.
For readers, a useful response is to look for evidence that bears directly on the book being considered. A sample can show whether the treatment of a subject is useful; a review can describe a reader’s experience; an author can explain the research or drafting behind the work. None provides universal certainty, but each is more closely connected to a purchasing decision than an account’s verified identity alone.
The defensible conclusion is that ebook production grew sharply and Amazon uses several mechanisms to manage it, including content review. How successfully those mechanisms establish provenance and protect readers is a separate question that the release counts and policy descriptions do not answer.
Written by the person on the byline, who did the research and stands behind every line. How our articles are made
