@jik I don't know of any studies, but the problem with their "invisible" product is that it's only invisible if it has n inference (true or false) that the user is human. So it reverts to the same accessibility challenges as the others in a portion of visits.