It's a question publishers are increasingly having to consider in the age of generative AI.

A post circulating on Threads recently reported comments from the Black Women in Publishing event suggesting that editors are considering asking authors for their last five drafts as a way of establishing provenance, rather than running manuscripts through unreliable AI detectors.

The idea has sparked an interesting discussion in the author community:
What exactly constitutes a draft?

Traditionally, we tend to think of a first draft as a completed, rough version of a manuscript. Subsequent drafts might tackle story, character, theme and pacing, incorporate feedback from editors or beta readers, and eventually move toward a final polish.

Under that model, asking an author for five drafts seems reasonable. Many writers will have separate files documenting exactly that progression. But not everyone writes that way.

Some authors edit as they go. By the time they reach "The End," parts of the manuscript may already have been rewritten numerous times.

Others write chapters in separate files, with Chapter 3 perhaps going through twelve versions while Chapter 15 barely changes at all. Some continually work in a single document.

So if a "draft" means a complete version of a manuscript, how are these authors supposed to produce five of them?

Google Docs and Scrivener can preserve extensive histories of a manuscript's development, which could provide excellent evidence of provenance. But even that doesn't necessarily produce five distinct, completed drafts.

Let’s face it, the bulk of the work exists within the creation of the first draft. Perhaps what publishers actually need is a record of the manuscript's development.

A snapshot every 20,000 words; an early chapter beside its heavily revised version; version history; editor and beta-reader notes; research files; outlines.

The evidence could look completely different from one writer to another while still demonstrating the same thing: this story developed over time.

Before generative AI, nobody particularly cared what went into creating that first draft. Now publishers may find themselves having to look behind the curtain. And when they do, they're going to discover just how differently authors actually work.

Asking for previous drafts sounds like an ethical and sensible alternative to running someone's manuscript through an AI detector and potentially making a false accusation. But requiring a fixed number of traditional drafts could inadvertently disadvantage writers whose creative process doesn't produce them.

Writing is art. Writers still need the freedom to use the process that works for them. Hopefully, as publishing figures out how to establish provenance, the question won't be:

"Can you show us your five drafts?"

but rather:

"Can you show us how this book came to be?"

posted by Sandra L. Rostirolla on August, 19