The chart circulating among self-published authors this year comes not from a press release, but from a working paper, and it shows something specific: how many new e-books Amazon has added to its store each month, tracking until the end of 2025 when its generative AI tools become widely available. The line does not go up. It increases about three times.
What the data actually shows
Economists Imke Reimers and Joel Waldfogel, a National Bureau of Economic Research working paper first published in January 2026 and revised in May, used Amazon release data to track monthly new e-book volume. Their findings are summarized NBER Digest: monthly new e-book releases averaged close to 100,000 in 2020-2022 and reached more than 300,000 per month at the end of 2025. That’s a monthly rate that has nearly tripled over several years, with no gap between before and after ChatGPT’s launch, but with a fairly close consumer following. AI writing tools, which researchers approach as a key driver of generative artificial intelligence.
What triple growth actually means for the storefront
Three hundred thousand new titles per month is not a marginal increase for a retailer to absorb. It’s a volume that changes what “self-publishing” means as a category, from a slower flow of individually written manuscripts to a flow that includes output alongside a rapidly assembled body of true single-author work that no single writer can match. Built for earlier, slower speeds, Amazon’s own catalog infrastructure wasn’t designed around the assumption that any given account could attempt to publish dozens of titles in a short window.
What Amazon actually changed
A specific answer confirmed by the company Publishers WeeklyIt was a September 2023 restriction that limited authors to uploading three new titles per day through Kindle Direct Publishing. This limits the speed of publication rather than verifying how any individual manuscript is produced, and it was adopted specifically in response to a wave of low-effort AI-generated content flooding the store faster than any review process could keep up.
What identity verification is available and what it verifies
Separately, the KDP own help documentation While Amazon hasn’t published the specific criteria that triggers it, the company does confirm that some accounts require proof of identity, a government ID, and a selfie. The widely repeated claim that it automatically kicks in once the account passes a certain number of published titles cannot be confirmed anywhere in Amazon’s own documentation or in the trade coverage of Publishers Weekly or the Authors Guild, and is best regarded as an unconfirmed figure that is thrown around alongside the real policy rather than the actual policy itself. What’s approved is narrower: Amazon verifies who a person is in some cases and regulates how fast anyone can post. No check reads a single word of what is actually inside the downloaded books.
The same nonsense as everywhere else in this story
This is a familiar form. The trade body’s authorship label verifies memberships, not manuscripts. The retailer’s disclosure policy checks the cover text, not whether the text is correct. Vitrin’s identity verification verifies that the person behind the account is the person, not what they actually wrote. Each of the publishers is checking AI-generated content this year, checking for something adjacent to an actual question, that specific text written by a person, because it’s something none of these systems are built to answer at the volume they have to work with right now.
What the numbers actually decide
A tripling of monthly e-book volume does not prove that any particular book was written by AI, and neither Amazon’s download limit nor its identity verification is designed to make that determination. The resolution of the numbers is narrower and more useful: the speed of production has changed enough and fast enough that systems built to handle the slower self-publishing pipeline are now actually doing better than the volume going through it, and the corrections applied so far drive the rate and the applicant, not the post itself.
What is really left to the readers
None of this gives a reader browsing a self-publishing showcase a reliable way to tell which of the three hundred thousand new titles a month have been written by someone sitting with material for months, and which have been put together at a speed that excludes it. Amazon’s cap slows down; verification of his identity confirms the applicant; none act as quality or authorship filters, and are not designed for either. This prompts readers to do the same thing with the honor code label and the disclosure discussion elsewhere in the publication this year: it does not currently perform a system-level check to filter by relying on reviews, sample chapters, and the author’s existing reputation.
This is not a criticism of Amazon specifically. No retailer on this scale has built a system that reads manuscripts before listing them, and none of the amendments proposed anywhere in the publication this year, the honor code label, the disclosure standard, the ID check, the daily cap actually do. The increase in volume didn’t just add more books to the shelf. It made it clear how many of those already sitting there were not screened in the way readers quietly thought.
For a reader trying to decide what to trust on this basis, the honest answer is that the volume of new titles has already outstripped any system set up to guarantee them, and no cap or ID check currently in place changes that math.
Whatever comes next, whether it’s better detection tools, tighter caps, or some version of honor code labeling being tested elsewhere in publishing, is likely to come before, rather than after, the volume challenge this story answers, exemplified at every corner.
Until then, the gap between how many books are published and how many are meaningfully peer-reviewed is the real story, and one that existing policy instruments are not built to close.
