The label is small enough to sit on the lower third of the back cover, about the size of an ISBN barcode, and says three things: “Society of Authors,” “Human Author,” and a logo mark that looks from a distance like any other publishing seal.
The proof is not shown to readers who pick up the book. They are shown a promise by the man whose name is on the cover that no generative artificial intelligence system has written the words inside.
What the label actually checks for
The Human Author schemaFounded by the Society of Authors of Great Britain in association with the Authors Guild of America at the London Book Fair in March 2026, it is self-proclaimed work. An author requesting a logo signs a license agreement stating that the text is written without generative artificial intelligence, although the scheme allows for aids such as spell-checking, brainstorming aids and software descriptions. It’s free for SoA members, limited to text-based work for now, and open to titles published from 2020 onwards.
The society does not read the manuscript before giving the label. His own Frequently asked questions is direct about this: “The SoA cannot make individual recommendations as to whether a work meets the criteria. This is an assessment that must be made individually by you as the author.” What the society confirms within 24-48 hours is that the applicant is a paying member. Claims regarding the manuscript itself are accepted on trust.
Applicable part and non-applicable part
The company owns the trademark on the logo, and this property is where its executive power really resides. If an author or publisher prints the Human Author mark without registering for the scheme, this is a trademark infringement that the Society can prosecute. If an author registers, signs a declaration, and the manuscript turns out to be substantially AI-generated, there is no equivalent mechanism. The integrity of any individual label rests entirely on the integrity of the person applying for it, and the scheme appears to have been deliberately set up this way: checking manuscripts for thousands of members would never be operationally realistic, so the trade-off was to protect the brand rather than policing the content beneath it.
Why did the authors want this?
The scheme comes amid figures the Society cites from its membership: 86% of authors say generative AI is already reducing their earnings, and 57% say they no longer see their writing career as sustainable. In contrast, a label that cannot verify content but can communicate intent is still considered by many members to be worth having. Novelist Tracy Chevalier introduced the scheme at the launch, and children’s author Malorie Blackman linked the motivation to craft rather than conformity: “The Human Author scheme seeks to emphasize imagination, commitment, craft and focus to produce stories and books that everyone can enjoy.”
The timing turned out to be more pressing than its organizers had planned. Ten days after the scheme was launched, Hachette was pulled Shy Girla horror novel, canceled in the US after its UK release AI-authorship claims — a case where it is difficult to hypothesize the code of honor nature of any claim of authorship. The label introduced to this environment was, among other things, an early answer to a question readers were already beginning to ask: how could anyone really know?
A narrower version on the other side of the Atlantic
Authors Guild of America a parallel version scheme and the difference between the two is instructive. The American program verifies the identity of the applicant using an identity verification service, but verifies the content of the manuscript like its UK counterpart. Two national trade bodies, working independently, reached the same limit: verify who asked what, not who wrote what. This approximation shows that the limitation is not an inherent control over an organization’s resources, but a structural fact of what a membership body can actually verify and only require its members to declare.
What does it cost to have your name on it
Self-declaration schemes are quickly dismissed as toothless, and narrowly speaking, no one checks the manuscript. But a label is not just a technical claim about text; A reputation developed under the name of a professional body whose credibility depends on the members not habitually lying. An author who subsequently registers a substantially AI-generated book does not simply risk correction. They risk becoming the model everyone cites the next time someone argues that the scheme won’t work – one that the Society is interested in informally valuing, like the reputation Shy Girl created for its author days after the accusation was made. The label does not need a verification system to carry fines. As Ballard’s work demonstrates well before any trade body intervenes, public attention is self-sustaining.
This does not equate to validation of an honor code. It’s betting that public oversight will do the enforcement that a membership organization can’t afford to do. The success of this bet depends entirely on how often everyone checks, and so far the checking has come from outside reporters rather than from the scheme itself.
What is the tag really for?
None of this makes the label meaningless, but it does mean that it measures something different than what its name implies. It doesn’t measure whether an AI is writing a book, it’s willing to sign a document, saying the author didn’t. For most authors, in most cases, this distinction won’t matter because most authors don’t lie about how they write their books. What the label actually formalizes is something that publishing always does without saying out loud: the possibility that the name on the cover corresponds to the person who wrote its contents.
The Human Author mark imprints this assumption on the record, in case some readers need to express more than is probably assumed.
