diff --git a/pkg/agent/anycast_linux.go b/pkg/agent/anycast_linux.go
index 4aaf403..dc0f907 100644
--- a/pkg/agent/anycast_linux.go
+++ b/pkg/agent/anycast_linux.go
@@ -130,7 +130,7 @@ func (r *AnycastReconciler) computeDesired() map[string]anycastTarget {
 		r.Store.Snapshot(),
 		func(ns, name string) bool {
 			pod, ok := r.Pods.Get(ns, name)
-			return ok && podReady(pod)
+			return ok && podAnycastEligible(pod)
 		},
 		func(s string) { r.Logger.Warn(s) },
 	)
diff --git a/pkg/agent/podinfo.go b/pkg/agent/podinfo.go
index ceb1cb7..6cd67ff 100644
--- a/pkg/agent/podinfo.go
+++ b/pkg/agent/podinfo.go
@@ -28,6 +28,16 @@ func podReady(pod *corev1.Pod) bool {
 	return false
 }
 
+// podAnycastEligible reports whether a pod should contribute its IP as a
+// nexthop for its anycast IPs. A pod is eligible when it is Ready AND not
+// being deleted. Once the apiserver sets DeletionTimestamp, kubelet has
+// started teardown — kube-proxy will keep routing for terminationGracePeriod
+// but the pod is on the way out; we should withdraw the nexthop immediately
+// so BGP shifts traffic to a sibling before the pod actually exits.
+func podAnycastEligible(pod *corev1.Pod) bool {
+	return pod.DeletionTimestamp == nil && podReady(pod)
+}
+
 // PodCache exposes a Get(ns, name) lookup against a node-scoped Pod
 // informer. ADD/DEL handlers consult it to read annotations + labels for
 // IPAM and (later) NetworkPolicy. Callers can subscribe to Ready
@@ -58,7 +68,7 @@ func StartPodInformer(ctx context.Context, cfg *rest.Config, node string, logger
 
 	_, _ = inf.AddEventHandler(cache.ResourceEventHandlerFuncs{
 		AddFunc: func(obj interface{}) {
-			if pod, ok := obj.(*corev1.Pod); ok && podReady(pod) {
+			if pod, ok := obj.(*corev1.Pod); ok && podAnycastEligible(pod) {
 				pc.fireReady()
 			}
 		},
@@ -68,7 +78,10 @@ func StartPodInformer(ctx context.Context, cfg *rest.Config, node string, logger
 			if oldP == nil || newP == nil {
 				return
 			}
-			if podReady(oldP) != podReady(newP) {
+			// Fire on Ready transition OR DeletionTimestamp transition.
+			// The latter catches "pod was Ready, now being deleted" so the
+			// reconciler withdraws the nexthop before the pod actually exits.
+			if podAnycastEligible(oldP) != podAnycastEligible(newP) {
 				pc.fireReady()
 			}
 		},
diff --git a/pkg/agent/podinfo_test.go b/pkg/agent/podinfo_test.go
new file mode 100644
index 0000000..abd0fa1
--- /dev/null
+++ b/pkg/agent/podinfo_test.go
@@ -0,0 +1,46 @@
+package agent
+
+import (
+	"testing"
+
+	corev1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+)
+
+func readyPod(deletionTimestamp *metav1.Time) *corev1.Pod {
+	return &corev1.Pod{
+		ObjectMeta: metav1.ObjectMeta{DeletionTimestamp: deletionTimestamp},
+		Status: corev1.PodStatus{
+			Conditions: []corev1.PodCondition{
+				{Type: corev1.PodReady, Status: corev1.ConditionTrue},
+			},
+		},
+	}
+}
+
+func TestPodAnycastEligible(t *testing.T) {
+	now := metav1.Now()
+	cases := []struct {
+		name string
+		pod  *corev1.Pod
+		want bool
+	}{
+		{"ready, not deleting", readyPod(nil), true},
+		{"ready, but deleting", readyPod(&now), false},
+		{
+			"not ready, not deleting",
+			&corev1.Pod{Status: corev1.PodStatus{Conditions: []corev1.PodCondition{
+				{Type: corev1.PodReady, Status: corev1.ConditionFalse},
+			}}},
+			false,
+		},
+		{"no conditions, not deleting", &corev1.Pod{}, false},
+	}
+	for _, c := range cases {
+		t.Run(c.name, func(t *testing.T) {
+			if got := podAnycastEligible(c.pod); got != c.want {
+				t.Fatalf("got %v want %v", got, c.want)
+			}
+		})
+	}
+}