Skip to main content
Submitted by admin on
Aggregate 核验溯源
Body
AI traffic is not one thing, and a policy which treats it as one thing gets the answer wrong in both directions. Here the decision is written down, class by class: the crawlers which harvest sites to train models are turned away at the edge before a single line of Drupal or Backdrop runs, while AI search indexers, and the fetch a real person triggers by asking an assistant, stay open and rate-limited per vendor, not per address, because one prompt fans out across dozens of addresses. Every layer ships with the platform and runs on your own server; nothing is rented, and no rule lives somewhere you have no login for. Changing it for one site is one line in a file you own, live in about two minutes, deleted to restore the defaults; and your own policy goes out at /llms.txt straight from the site's files folder. The honest part is the best part: a user-agent can be forged, the maps fail open on purpose, and the docs say where each shield stops and the next one begins.
Domain Tag
drupal
Source Name
Drupal.org Planet