At storytell, we added some measures to make the response based on facts that appear in the content.
This would make our result more factual and reliable than the output of ChatGPT. However, it will also make the result less creative than ChatGPT. You might get answers like `the content didn't mention XYZ, so there is no response to provide` when trying to ask open-ended questions.
This is a ultimate tradeoff that we need to make when building AI. We make the choice of prioritizing reliability because we view it as crucial to deliver long-term value to users.
In the future, we might add some options for users to choose mode. We are also working on feature to allow Storytell to return more results based on multiple pages (not just current page).