Google published research on how they're using Cluster Detection (behavioral patterns) and Synthetic Content Rating (semantic patterns) to purge scaled AI slop from YouTube.

Seems like they don't want to accidentally take down human creators who may or may not use AI tools, but it's good to keep in mind and maybe do things to avoid sending the wrong kinds of signals:

• Avoid templated, repetitive title and description structures
• Have a realistic human publishing cadence, don't batch upload/publish huge volumes of videos using scripts or API calls
• Don't use a third-party video generator other other automation service to manage your uploads
• If you use AI for voiceovers or visuals, make sure your video also contains original human commentary, specific context, a "human touch" editing style, etc

The research paper only talks about online video platforms but it's not a stretch to consider how Google might apply this type of semantic classification/cluster analysis to web content too. Low-effort AI-based content scaling: just don't!

Here's a link to the abstract where you can download the PDF report