Libreon isn’t trying to determine whether a manuscript “looks like AI.” We’re trying to establish whether a human wrote it.
That means reviewing author documents and how they changed over time, and any other artifacts of the creative process that can be shared, along with an attestation of AI usage in that process.
No single draft, document, answer, or detection score decides the result. We look at the totality of all evidence.
The submission form is in-development. For now contact us on our Chat Server.
LIBREON VERIFIED Human-Crafted is a retrospective assessment of an author and a manuscript. To earn it, the author provides evidence that allows Libreon to determine whether the work meets the Human-Crafted Standard. The assessment includes:
Libreon Journeys are a guided manuscript-development system. Authors work through the manuscript with mentors, mentor others, and complete milestones along the way. By the time a manuscript reaches its final milestone, the process has proven its provenance.
Human-Crafted does not mean authors have to avoid modern writing tools simply because those tools incorporate AI. Authors may use tools in their process, so long as that use remains within the Libreon Human-Crafted Standard and it adheres to our Generative AI Policy, which states:
Every word in your writing should come from your own hands—and similarly, every brushstroke in your artwork.
The important distinction is authorship: did the tool help the author, or did the tool do the work? Helping can be permissible. Supplying manuscript language isn’t.
What isn’t compatible is generated language being presented as the author’s own writing, such as:
A VERIFIED assessment begins with the author submitting evidence from the manuscript’s development and other required information such as attestations from contributors.
A Libreon assessor reviews those materials and then meets with the author for a confidential, recorded video interview. They’ll talk about the book, how it developed, and the author’s creative process.
This isn’t a literary exam. There are no standard questions and no “correct” answers. We are not looking for polished work—we actually want to see messy early drafts and how the manuscript changed along the way.
Possible Outcomes
A verification assessment has three possible states:
An inconclusive result does not mean that a work was generated by AI. It means Libreon could not establish sufficient provenance to issue its certification.
Libreon does not publish a list of works or works receiving inconclusive results. We do publicly identify works that successfully receive certification, as well as those that are revoked.
An inconclusive outcome comes with an explanation of why. The author may make one appeal. An appeal does not guarantee a different result.
The assessment fee is refunded only if Libreon does not return an outcome within 60 days. An inconclusive result is not grounds for a refund, nor is a certification that is later suspended or revoked.
Once certified, a work may display the appropriate Libreon certification seal according to Libreon’s seal-use guidelines. Each seal includes a QR code that readers, booksellers, publishers, agents, and others can use to confirm the work’s current certification status.
Certification remains conditional on compliance with the Libreon Human-Crafted Standard and certification agreement. Libreon may revoke a certification if material information was false, omitted, or later shown to undermine the basis on which certification was granted.
After a revocation, the author or rights holder must stop using the certification mark and, where reasonably possible, remove it from digital materials. Physical materials manufactured before the revocation do not have to be destroyed solely for that reason, but depending on the situation, Libreon may require the seal on existing physical materials be covered with a sticker, at the author or rights holder’s cost.