If you have a book or other writing online, two separate decisions about it came up this week. One is whether AI companies can train on it, which means using it to teach their systems. The other is whether you let AI read it aloud. They run on different timelines, so I'd make them one at a time, starting with the one in the news.
What changed this week
As of October 2026, The Next Web reports that Google, Microsoft, OpenAI and Anthropic each say they will honor one shared robots.txt line for opting out of AI training. Robots.txt is the plain text file on a website that tells automated visitors which pages to leave alone. The line is a policy token called __usage_ai-training. You add a group for it to your robots.txt file and list the paths you don't want used for training.
The idea comes from an individual IETF draft dated October 2026. Its only listed author is an Anthropic engineer. The IETF has not endorsed it. That means it has no formal standing in the standards process.
The four companies are not starting at the same time. Only Microsoft has given an exact start date, February 1, 2027. Microsoft also says the new preference replaces the NoArchive and NoCache tags it has used since 2023. Google aims for 2027. Its crawler documentation says early 2027. OpenAI says "in the coming months." Anthropic has started work and gave no date. Google has a blog post on the change.
Those details limit what the opt-out can do. The draft says the token states a preference and is "not a directive for crawling behavior." A site with no such opt-out group has stated no preference. Private agreements between a site and an AI company can override the token. Microsoft told the reporter that its new preferences do not establish licensing or payment arrangements.
For an author, the practical part is small. Adding the opt-out group takes one edit to a text file and puts your preference on the record. That's all it does. Say you have a site with sample chapters, transcripts or a full-text blog. Decide which paths hold the material you care about and list those. Anything you've licensed privately may fall under that agreement instead.
Reading your work aloud is a separate thing
That covers the first decision, which is about protecting your writing. The second is about putting it to work. Opting out is about what you don't want AI companies to take from your work. Narration would be something you make yourself, with AI, from that same work. The one report I found comes from Inkfluence. It sells AI audiobook tools. In a post dated August 2026 it analyzed 8,358 chapters across 872 books and 529 authors on its own platform. Chapters narrated per month went from 3 in November 2025 to 2,914 in July 2026. 99% of those narration jobs were completed cleanly. The same post says fewer than 1% of their registered users have made an audiobook. Those numbers show that more authors are trying narration and that the jobs mostly finish without trouble.
The post doesn't say how the audiobooks sold or whether anyone listened to them. It also calls its roughly ten-minute average chapter length a close estimate. It lists where the files can be deployed. Google Play Books and Kobo accept AI narration with disclosure. Audible's ACX requires human narration. The post warns that these policies can change. Check each platform before you plan.
Why I'd decide about them separately
Put the two together and the difference is what happens when you get it wrong. The opt-out is a statement about the future that no company fully honors yet. And you can change it by editing a text file. Narration produces a product. It has a platform policy to satisfy and a listener who hears the result. So if there is a mistake, it is public.
One more limit applies to both sources. Neither the news report nor the Inkfluence post gives listener results for AI-narrated books, and neither says anything about books in languages other than English. If your catalog includes one, these sources don't give you much.
Try one chapter
With that said, you can start without committing to much. A single chapter makes a workable test. Pick one from a book whose rights you hold. A chapter with a table or coined terms in it is best. Have it narrated and then listen to it. Then open your robots.txt and write down the paths you would list. Comparing how each step felt will show you why I'd treat them as two decisions.