Is it? Couldn't the backend (or even a human attacker) just type the credentials you provide into the real login page, giving you the "tap yes" push notification just the same?
But now that I think about it, it would make sense to combine new device notification with push-notice 2FA for exactly that reason, since you've got a push channel that takes a confirmation already, flag unexpected devices in that channel as well and it becomes much more secure.
An attacker can play the legitimate WebAuthn request from the real site, which will (statistically certain) be nonsense if played by their phishing site.
Or they make their own request, which doesn't help them because it's not valid on the real site they want to sign into so it's pointless.