Kashmir Hill Finds AI-Generated Biography of Herself on Amazon

Hill's first-person account says the book combined public information about her life with generic prose and invented details. It incorrectly described a coffee-making ritual, claimed access to her childhood diary, and suggested interviews with people who knew her. Hill said she had no involvement in the book.

The author profile used a stock photograph and described Miller as an experienced biographer. The same account had published 10 biographies of journalists in one week. Hill summarized the episode in a public LinkedIn post, calling the biography AI-generated and linking to her investigation.

After Hill left an Amazon review asking Miller to contact her, the biography disappeared, along with the other titles associated with the account. The available sources do not establish who operated the account or whether Amazon, the publisher, or another party removed the books.

Hill's investigation also documented a separate, identifiable publisher, retired cybersecurity consultant Bill Johns. He told Hill that he used ChatGPT in a workflow that produced 445 nonfiction books for Amazon and targeted 10 new titles per week, which he described as the platform's limit.

That example is distinct from the unknown Miller account, but it shows why inexpensive generation and print-on-demand distribution can create large catalogs quickly. Hill reported that Amazon asks publishers to disclose AI use during submission but does not show that disclosure to buyers. Amazon told her it invests in enforcing its guidelines and removes books that do not comply.

The incident is a provenance failure, not just a writing-quality problem. A disclosure that AI assisted with a book would not verify the identity of its author, the consent of its subject, or the accuracy of its biographical claims.

For publishing and content-generation systems, stronger controls include verified publisher identities, subject-complaint channels, evidence requirements for purported interviews or private records, rate-limit review for high-volume accounts, and clear buyer-facing disclosure. Those measures address the accountable origin of a work rather than relying on style-based AI detection alone.