The question itself (ie if the user is under 13) doesn't matter. Already for the current legislation there's nothing stopping the device owner from intentionally lying about the age. So really this entire exercise is about providing a standardized means of control over filtering, thus my observation that the proposed measure is both backwards and overly limited in scope.

The software on the device can do whatever it would like with the signal it receives, including consulting the user account metadata for declared age if the device owner so desires.