Put every page’s permissions in one place
Why now: The shared training signal makes it useful to organize what you want for each page separately from how providers implement those choices.
Prompted by today's lead: Shared AI-training opt-out gains four major backers
What you'll need
- Non-sensitive notes naming website pages and your intended access and AI-training preferences.
Steps
- Collect the page names and permission preferences you already have in writing.
- Paste the notes into the prompt to create a plain-text register.
- Check each entry against your notes; keep the register as a statement of intent, not proof of configured controls.
The prompt
Organize these website notes into a content-permissions register. Use plain text with no Markdown formatting and fewer than 150 words. Write one line per named page, using these fields: Page; Public access preference; AI-training preference. Preserve the stated preferences without interpreting public access as permission for AI training. Write 'unspecified' for any missing preference. Do not recommend technical settings or claim that any preference is enforced. Notes: [paste your website notes]
What you should get
Three plain-text lines, one per page, with all three requested fields; the old-map index’s training preference marked unspecified; no technical settings or enforcement claims; fewer than 150 words.
Tested example: Tested on sample input
Input
Our volunteer local-history website has three pages. The walking-tour page should be readable by everyone, but we do not want it used for AI training. The oral-history transcripts should stay accessible only to volunteers, and should not be used for training. The old-map index can be public. We have not discussed a training preference for that index.
Output
Walking-tour page; readable by everyone; not for AI training Oral-history transcripts; accessible only to volunteers; not for AI training Old-map index; public; unspecified
How we tested it: How we checked this: an AI model ran this prompt on the sample input on October 30, 2026, and a separate model call checked the result against what the Move promises.