I did some other "AI" writer more than 5 years ago and I had to grade it a C- to a D-, almost unusable. Things have changed a lot since then. This isn't a 1st generation product and it's nice to give feedback to and get responses from Ioannis, the purveyor. We had a back and forth over some of the pain points where editing is the largest painpoint of this product if you want it published. A bunch of things have been fixed and there is an analyzer of the prompt which grades it to 100. The 6000 character prompt limit is a problem if your book outline/toc is 9-12+ chapters. Human model does well for words over a certain amount. There should be another section with its own limit for the toc/outline because of the 6000 character limit. https://www.youbooks.com/roadmap shows citation concerns that are being worked on. For example, "RAG (retrieval-augmented generation) hallucinations are instances where a large language model (LLM) generates incorrect or fabricated information even when using retrieved documents for context." Giving this feedback helps them to consider Chicago footnote style citations/references and bibliography page automation instead of simply getting an export of URLs for us to edit. The TOS of websites like Reddit does not allow for scraping for training. But, you can use sources like sourcely.net (site is down as of this writing but I've subscribed before). They have a future vision for Advanced Publishing, among 3 others, but they can simply offer an API key for developers. If not, allowing for a 3rd party addon marketplace for post-publishing is also good but more complex under contractual agreement. They fixed the header problem which makes it easier for you to edit portions of your deliverable. If things are modularlized, then you can do post-publishing such as add chapters. If you ask for doctorate level content, other AI models find it as such. I've had it make more than 50+ books. The exact number is known by OCR into Google Sheets and categorized for genre and subgenre. It's going to take you some time to nail down your own process. It's capable of making fiction because the prompt box takes in toc/outline. But you have to know what to parameterize. "Use Freytag's Pyramid framework (for the plot), temperature (randomness) average, base it on the real life story of..., use this outline/toc, use a narrative structure with dialogue, etc." I've had it write me a language learning book and it's in the A range as a product. It does disregard your toc suggestions but now, you can review and modify it. Ethics-wise, it is biased mainly in your favor and the RAG guardrails of the AI model. But, all AI models are sophisticated mimicry which precludes them from true creativity, productivity, value creation, and human connection. Human review is required based on the subject matter as it can side with misinformed consensus instead. An example of this is for me as a creationist is macro-evolution. The AI models used does do RAG in varying degrees but the conclusions are invalid. LLMs are part statistical guesswork on the contents of the sentence and the next sentence in a paragraph. It does logic checks more with the outline and its consulting the Named Entity Recognition (NER) database for things like country names. Most mutations are harmful and cannot overcome the frequency required for the low number of beneficial mutations for evolution to be true. Protein folding can't be explained either by evolution, etc. So, it mimics reasoning but it can still hallucinate even with RAG. However, this platform also has huge potential as fiction is more lucrative but the manual post-production editing is very time consuming. It is only as good as your understanding of prompt engineering, and that took me several months through trial and error. I wish Ioannis/Youbook much success. Great product if your goal is to have lifetime ghost writing assistance.