Rarely, when published unedited. Generative systems favour sources offering something they cannot synthesise themselves — original data, named expertise, specific evidence. Content generated from the same general knowledge adds no reason to cite it.
Using AI for research and drafting support is not the problem; publishing unedited output is.
Original evidence is the differentiator: your own data, your own testing, your own named practitioners.
Fact-check every statistic before publishing regardless of who drafted it.