I would guess these narrow docs bots probably perform worse than ChatGPT et al in 'search' mode