AI for Authors

What should I check before uploading my manuscript to an AI tool?

By the WriteLoom editorial teamUpdated 2026-06-03
Key facts
  • Check whether the tool trains its models on your uploaded text.
  • Check the data retention period and where the files are stored.
  • Check access controls — who at the company can read your manuscript.
  • Check that you can export your work in a standard format.
  • Check that you can delete your data fully and verify the deletion.
Direct answer

Before uploading an unpublished manuscript, check five things: whether the tool trains its models on your text, how long it retains your files, who can access them, whether you can export your work in a standard format, and whether you can fully delete your data. Unpublished work is your most valuable asset, and the terms of service — not the marketing page — tell you how a tool treats it. If training is on by default or deletion is unclear, treat that as a reason to look elsewhere.

Chapter i·Why it matters

An unpublished manuscript that is used for training, retained indefinitely, or locked inside a tool you cannot export from is a real and often irreversible loss. Authors rarely read the terms before pasting in 90,000 words, and by then the upload has happened. Running the five checks first costs a few minutes and protects the one thing in your career that is genuinely irreplaceable: the work itself, before anyone else has seen it.

Chapter ii·What to include

  • A training check: is your text used to train models, and can you opt out?
  • A retention check: how long are files kept, and where?
  • An access check: who at the vendor can read your content?
  • An export check: can you get your work back in a usable format?
  • A deletion check: can you remove your data fully and confirm it?
  • A read of the actual terms of service, not just the feature page.

Chapter iii·Example

Before trying a new editing tool, an author searches its terms for the words "train," "retain," and "delete." She finds that uploads are used for model training unless she changes a buried setting, and that deletion is not guaranteed. She switches the setting off, tests the tool with a non-sensitive sample chapter, and keeps her full manuscript on a vendor whose terms promise no training and one-click deletion.

In WriteLoom

WriteLoom does not train on your manuscripts, lets you export your work anytime, and supports full deletion — your unpublished writing stays yours.

Read our data promise