A: Unfortunately, it will not work the way you intend it to, i.e. you cannot instruct the AI on how to extract and how to handle the content. You are absically right in that the AI decides for itself on what to pull (which seems random to you). This is done using a similarity search and context injection (commonly known as RAG). The problem with RAG is that it does not work well with specific instructions like the ones you mentioned.

However, results should not be "generic". We have ran hundreds of evals of youbooks generated books against books generated by other AI apps and they come way ahead in depth. This is no wonder, as youbooks scraped hundreds of pages to put your book together and can write very long books (10x bigger than most competitors). Other things that will not work include referencing page numbers or files names.

If your book feels generic, try increasing the word count. Youbooks will budget subject depth based on word count. If you target 30K words, you will not get an in-depth book. There is no word budget for it, assuming a certain target density and repetition. Also, try writing the TOC yourself (inject it in the prompt) to avoid having word budget wasted in topics you don't want covered. Another trick you can try is to target a higher word count than the one you have in mind, then ask Youbooks to trim repetition heavilly and increase the Edit Level to 3. Then, edit the produced book even more and remove content to make it more dense.

On citations, no, Youbooks does not include in-line citations because of the sheer amount of content scraped and researched. It's a trade-off that we made from the beginning to enable research on such a wide scale. We do plan to include file that will list all the urls scraped during writing the book, on a per section basis (not in-line). This is coming soon. You should be able to use that to check facts using an agent, but it will be quite an undertaking.