GRE is fundamentally a stateless encapsulation protocol: the encapsulating router simply wraps the passenger packet in a GRE header and an outer delivery IP header and forwards it based on ordinary IP routing table lookups toward the configured tunnel destination address. For that forwarding to succeed, the underlying (global or routing-instance) routing table on each endpoint must contain a valid, reachable route to the far-end tunnel address; without such a route, the tunnel interface will not come up or will be unable to deliver traffic, since GRE relies entirely on the existing IP infrastructure rather than any tunnel-specific signaling to establish reachability. Because GRE has no built-in state machine or session negotiation, it does not natively track whether the remote endpoint is alive - that requires an optional add-on such as GRE keepalives or BFD, neither of which is enabled by default; an operator must explicitly configure them if failure detection is required. GRE is also protocol-agnostic rather than IPv4-only: it can encapsulate IPv4, IPv6, MPLS, and other network-layer protocols as the passenger, which is one of its defining advantages over IP-in-IP tunneling.
Finally, because GRE keeps no per-session state about the peer, it cannot detect or report tunnel failures on its own, reinforcing why route reachability is the one true prerequisite for tunnel operation. Reference topics:
Junos Enterprise Routing - Tunneling, Understanding Generic Routing Encapsulation.