I prefer buttons because there's less risk of accidentally highjacking the gestures used by mobile screen-reader users to navigate.
Also buttons theoretically tell you what they do before they do it (when they're not just abstract icons). Gesture relies on cognitive load, which forces me to remember an action mapping. Which I do not. Especially since covid. Or i have to abandon my task to search for documentation, which is worse because it reduces the odds I'll compete the thing I was trying to do.