I vividly remember scanning 1000+ books into Delicious Library, which was vastly faster than entering them all completely manually, but something like this LLM based book scanning would be so much faster, particularly if it was wrapped into a real tool that did full lookups, etc (like Delicious Library did).

The possibility for hallucinations doesn’t worry me much - quickly verifying the results afterwards (plus the second layer of a lookup) would mostly solve that. I’d be more worried about not noticing something was missed, but good tool UX could probably solve that.