It's not end-to-end encrypted, nor authenticated for that matter. There's no substitute for DNSSEC, though like a VPN, DoH makes it easier to evade local or regional interventions. (And because I know DNSSEC triggers some peope, I'll point out myself that WebPKI is a proper end-to-end solution... mostly. Without DNSSEC acme verification techniques are less than ideal.)
Does anyone know how the automatic (non-system) built-in DoH in browsers selects a resolver? I assume they don't globally utilize the same DNS resolvers, as they would muck with GeoDNS; at least if those resolvers aren't doing the backend resolution in a way that preserves locality.
Keep in mind the audience making the decisions here is perfectly happy to require that geographical maps be censored to obscure sensitive locations. From their perspective, this is much the same thing.
It's all about truth and messaging.
DNS tells you where something is, no more and no less. When queried, it points to the map and says: "It's there, off you trot". DNS does not care what is there, it is just a guide.
DNS works over port 53/udp+tcp and could be fiddled with in transit. There is a variety of DNS called DNSSEC which, with care, can be trusted to be accurate (for your given value of accurate).
Messaging: As I said, port 53/udp+tcp could be fiddled with if DNSSEC is not in play. There is also DNS over HTTPS or DoH. Now the messaging channel is encrypted. At this point you have the same confidence in security as you do when looking at https web sites.
Now, you can obviously use DoH DNS from anywhere in the world and it will look like ordinary web traffic. Most browsers are already doing this.
Messing with DNS now in Europe is basically fiddling whilst Rome burns - bloody stupid and couterproductive. Hamstringing a service such as Quad9 is at best ridiculous.