In a support discussion titled “Embeddings from limited category?”, Jagster asks whether AI embeddings are generated for topics in categories not visible to all users—specifically, for a curated category designed for AI consumption, where content is streamlined for better Q&A accuracy. Jagster notes that while muting categories or tagging posts might hide them from users, it doesn’t block Google or embeddings. Falco clarifies that AI search respects the same visibility rules as keyword search, meaning if a topic is accessible to a mentionable AI persona (via group membership), it can be returned in AI search results. Jagster confirms that AI search does find such topics when the bot has access, but questions whether the embeddings model itself generates database entries from restricted categories. Falco confirms: yes, embeddings are generated for all topics regardless of visibility, meaning even hidden or curated categories contribute to the AI’s knowledge base. The core goal is to reduce noise and improve precision in AI-generated answers by curating content, though Jagster acknowledges possible gaps in understanding how embeddings, RAG, or GPTs function. The discussion touches on both technical behavior (embedding generation, visibility rules) and practical intent (optimizing AI performance through content curation).
Somekind rule of thumb is AI can’t see and operate if at least reading is harder than everybody. That is not true, if mentionable AI-persona is member of a group that sees such category.
But how are embeddings? Will those be builded even a category isn’t visible for everybody?
And the real issue I’m trying to solve out is a category where all posts are purpose made (I just can’t remember right writing form… curated?) for AIs, so those are actually copies of ”real” topics, but not 1:1.
There isn’t any secrets but I would like to keep googlebot out of it. Plus such content is a bit not so user friendly.
Surely I could mark those with special tag and mute it for users. Or/and mute whole category. It keeps most of users away, but I have a feelint that Category Navbar will show muted categories too; I may remember that totally wrong, though. But muting doesn’t keep Google out.
By my experience posts, and specially if that topic is longer are… how should I put it… are treated differently by GPT, kind of like OP would be more important, if you follow me? And I would like to clean away all noises that will be there. By that way I get more accurate Q&A content with better context and more precisly answers. Sure, the main issue may be my prompts, system and user, or OpenAI. But my topics aren’t that… straight and black&white as here, or generally in dev/coding communities.
Sure, I may lack important basic understandigs how embeds, RAG or even GPTs work and what I can expect. But that way I got way better answers than using same prompts targeting normal topics.
Basically I shrink important content of a topic to new topic with one post.
For AI search, we apply the same visibility checks as the keyword based search, so if those topics can be returned by the keyword search, they may also be returned by AI search.