Hi @User1619 -- Cam from the Intercom Support Engineering team here!
Fin uses a mix of AI models and LLM technology to understand customer questions, search for relevant answers within the Content Sources you’ve setup, and provide accurate answers to customers based on the data contained in those sources. I’ll note that it will call on different sources at different times in an effort to strike that all important balance between latency whilst generating an answer and providing the correct answer to the end-user.
There’s no single format or source that works best for Fin or is explicitly prioritised by Fin when constructing an answer. However, Snippets are a particularly effective way of allowing you/your teammates to quickly and easily add content for Fin that might not be publicly available - like internal notes, bug or issue details, or conversation replies - or emphasis data that may be contained in longer or more complex Articles (like an FAQ) to allow Fin to utilize the content more effectively.
If you’re noticing that there is an Article(s) not being used, you might consider looking into whether there’s any way to optimize the Articles content so it’s easier for Fin to consume (best-practice recommendations on this here).
I’d also highly recommend having a look at the details provided in the Fin answer debugger for this question (available when an AI generated answers is provided) to understand how Fin has answered the question, and suggest improvements on how Fin handles future queries. You can find more details on the Fin answer debugger here and here.
(NOTE: If you have Custom Answers enabled, these will always take precedence over AI Answers for the same question. If Fin doesn’t find a Custom Answer for a question, only then will it search your Help Center content to provide an AI-generated Answer to the end-user. Custom Answers are turned “Off” by default.)
Let me know how you go and if these details help answer your questions!