I read this technical article in QST September 1927 about second harmonics in Regen receivers:
https://archive.org/details/sim_qst_1927-09_11_9
It's an interesting topic and I have one kind of regen circuit where it is kind of implementable. I might try it some time.
Anyway most AM detection circuit have a non-linear response and are less sensitive at low input level (compared to a true linear response) than they are to high level inputs. Which aids in selectivity.
Also you can put a regen into direct conversion mode by having it slightly oscillate at the same frequency as an AM transmission and get more selectivity that way. However the regen needs to be very frequency stable for that.