Whether writers like it or not, generative AI has created a legitimate provenance problem. Publishers need some way of navigating it. My recent post on Instagram about publishers potentially asking authors for previous drafts as a way of establishing provenance generated hundreds of comments.

After reading through them, one thing became very clear: We all write books very differently.

Some writers had dozens of drafts. Others had one continuously evolving document. Some wrote each chapter separately. Others had notebooks, Post-it notes, research files, character sheets, discarded scenes, Google Docs, Scrivener snapshots and random paragraphs spread all over the place. Writers don't dictate when an idea hits, so they'll use what ever they have on hand to jot down the inspiration.

If a publisher asks to see the last five drafts of a manuscript, they're largely seeing what happened after the first draft was written.

Maybe the better question should be: How did that first draft come to exist? That's where the lion's share of the creative work happens.

Even better, maybe a discussion should be had. Maybe publishers should be asking writers:

- Why a character made a particular decision?
- What changed from the original idea?
- Which part of the story caused the most trouble?
- What scenes were cut and why?
- Did a character grow from incidental to largely taking over?

Writers who have lived and breathed their story know their characters inside out. They talk about them with passion and excitement. Of course, an author can still largely write their story, develop their characters and still have AI draft the prose.

The reality is, there is no perfect test. AI detectors are not the answer, neither is a one-size-fits-all send five drafts.

posted by Sandra L. Rostirolla on September, 01