Announcements
Is it documented anywhere or has anyone figured out what is going on behind the scenes of the "OOB" Generative Answers against SP? Is it just executing a SP search request via graph, then generating a response? I ask because I'm not real impressed with the performance and UX, and may just roll my own via calls to Power Automate, etc. but want to make sure I'm not over-assuming or missing something.
Hi @darylvogan,
It does a bit more than that. It's using the Graph API to make a delegated call in the user's context, then retrieves the top documents (hence 3MB size limit) to identify the relevant content snippets to summarize and generate an answer. Generative answers also add a few layers for moderation (e.g. harmful content) and validation (to try to catch and remove hallucatination).
Soon, for users eligible to the Semantic Index for Copilot, results will returned using this index, allowing for more complex queries and better results.
@HenryJammes Thanks for this answer. As a follow up question: After retrieving the most relevant documents how are the most relevant chunks then selected? Are the documents chunked on the spot and then embedded or is it something else?
Trying to figure out whether to start using copilot studio for sharepoint docs or stick to ingesting the docs and then creating a vector database.
Under review
Thank you for your reply! To ensure a great experience for everyone, your content is awaiting approval by our Community Managers. Please check back later.
Congratulations to our 2026 Super Users!
Congratulations to our 2025 community superstars!
These are the community rock stars!
Stay up to date on forum activity by subscribing.
Valantis 696
chiaraalina 171 Super User 2026 Season 1
Haque 122