A British startup called Books by People announced on 7 October 2026 that the UK Intellectual Property Office has approved its certification mark for books written by a human. The stamp looks like a thumbprint and is meant to sit on a book's cover. Publishers can use it to tell readers a title is human-authored, without any ban on books made with AI.
“We need to be quite careful that this doesn’t end up becoming a witch hunt.”
Tom West, CEO, Publishers’ Licensing Services, to WIRED · wired.com ↗
To earn it, an author submits a declaration, and the company runs the work through its own software and has staff review early drafts and research material for what it calls indicators of humanity. Its own materials add that titles are registered and checked against creation accounts, manuscript history and signature analysis. Publishers pay a flat membership fee plus a charge for each book they put forward. Five have signed up so far, and Galley Beggar Press says it intends to adopt the mark. Certified titles will appear in a public directory with a unique ID and QR code, and the status can be passed into retailer data. The first marked book is expected by the end of 2026.
Cofounder Esme Dennys describes the aim as giving readers a way to find human-written books while leaving those who want AI books free to read them. The company stresses that this is not AI detection, and says that when it is unsure it errs toward the author. That separates it from detection vendors such as Pangram, which claims 99.7% accuracy. Detection tools have a record of misfires: debut author Jerry Falade said software flagged his Nigerian-dialect prose, and Pangram scored his book 97% AI.
The scheme arrives after a run of 2026 publishing scandals over suspected AI use, including a reported $2m deal that was cancelled, a withdrawn novel and a disputed prize-winning story. Earlier honesty-based efforts such as Human Authored came first. Some critics argue that labels should mark AI-made work, not human work. Others, like author Ted Chiang, see a no-cheating policy as a statement of values even when enforcement is imperfect.
Reaction from inside the book trade is mixed. Sam Jordison of Galley Beggar Press, the publisher planning to adopt the mark, said readers would feel something close to revulsion on discovering that an AI text had passed itself off as human. Publishing figures also worry about the cost of getting it wrong, since a book wrongly doubted could harm an author's reputation, which is why the company says it leans toward the writer when a case is unclear. The reviewers look at process evidence rather than prose style alone, so a writer needs drafts and research notes to hand. Falade's case shows the risk the startup is positioning itself against: a tool read a dialect as machine output and attached a 97% score to a book he wrote himself. Whether the stamp wins readers' trust will depend on how the review is applied once the first certified titles reach shops.
Sources · 2