Generative AI Warning for Authors (Sample)

This sample client-facing language was developed by the AI Committee in June 2026. Members who wish to communicate their AI expectations to clients are welcome to use and adapt it for their own purposes.

Generative AI can be a useful tool, and we’re not telling you to avoid it. Brainstorm with it, work on an outline, do some research, or ask it where your draft is weak: these uses are all reasonable by most standards, as long as you’re using a locally hosted model or a paid one that will not use your work for training purposes.

Here’s what matters: beyond using an AI like a thesaurus, none of the words in your manuscript should come from one. This includes more than generating text from a prompt. You should also avoid using a tool like Claude or ChatGPT to edit your text directly: no “tighten this,” no handing over a chapter and keeping what comes back. The moment an AI touches your sentences, you face three problems:

Your reputation. Detectors like Pangram now flag text that an AI merely edited, not just wrote, with high accuracy. And one stray “Certainly! Here’s a revised version…” left in the document can sink your credibility in a single screenshot.

Accuracy. Even the most advanced AI models routinely invent facts, quotes, and sources with total confidence. In nonfiction, one fabrication that slips through represents a career-ending liability.

Copyright. The U.S. Copyright Office has made clear that AI-generated text can’t be protected by law. Let an AI write or rewrite parts of your book, and you muddy what you actually own.

Even “busywork” is risky. AI has hallucinated fake citations when asked to do nothing more than format an author’s existing references. Use dedicated tools like Zotero for citation management.

Bottom line: Make every change to your text by hand. Use AI to assist in research, planning, and organization, but never to edit or write the text itself. If you’re unsure whether something crosses the line, it’s safer not to do it.

One more thing: AI raises other concerns too, from confidentiality to how the models are trained to environmental cost. We’re setting those aside here. This warning is for authors who have already chosen to work with AI and need to understand potential risks.