diff --git a/libs/gorules/rule_benchmark_loop.go b/libs/gorules/rule_benchmark_loop.go
new file mode 100644
index 00000000000..50eac3ab494
--- /dev/null
+++ b/libs/gorules/rule_benchmark_loop.go
@@ -0,0 +1,15 @@
+package gorules
+
+import "github.com/quasilyte/go-ruleguard/dsl"
+
+// UseBenchmarkLoop detects classic benchmark loops over b.N and suggests
+// b.Loop() (Go 1.24+; the inlining regression was fixed in 1.26 so there's
+// no longer a reason to keep b.N-based loops).
+func UseBenchmarkLoop(m dsl.Matcher) {
+	m.Match(
+		`for $_ := range $b.N`,
+		`for range $b.N`,
+	).
+		Where(m["b"].Type.Is("*testing.B")).
+		Report(`Use 'for $b.Loop()' instead of looping over $b.N (Go 1.24+, performance-correct in 1.26+)`)
+}
diff --git a/libs/structs/structdiff/bench_test.go b/libs/structs/structdiff/bench_test.go
index f0406047835..08b386416ef 100644
--- a/libs/structs/structdiff/bench_test.go
+++ b/libs/structs/structdiff/bench_test.go
@@ -17,15 +17,13 @@ func bench(b *testing.B, job1, job2 string) {
 
 	total := 0
 
-	b.ResetTimer()
-	for range b.N {
+	for b.Loop() {
 		changes, err := GetStructDiff(&x, &y, nil)
 		if err != nil {
 			b.Fatalf("error: %s", err)
 		}
 		total += len(changes)
 	}
-	b.StopTimer()
 
 	b.Logf("Total: %d / %d", total, b.N)
 }
diff --git a/libs/structs/structtag/jsontag_test.go b/libs/structs/structtag/jsontag_test.go
index b45fd566ea1..810d66851b7 100644
--- a/libs/structs/structtag/jsontag_test.go
+++ b/libs/structs/structtag/jsontag_test.go
@@ -59,7 +59,7 @@ var benchTags = []JSONTag{
 }
 
 func BenchmarkJSONTagName(b *testing.B) {
-	for range b.N {
+	for b.Loop() {
 		for _, tag := range benchTags {
 			tag.Name()
 		}
@@ -67,7 +67,7 @@ func BenchmarkJSONTagName(b *testing.B) {
 }
 
 func BenchmarkJSONTagOmitEmpty(b *testing.B) {
-	for range b.N {
+	for b.Loop() {
 		for _, tag := range benchTags {
 			tag.OmitEmpty()
 		}
diff --git a/libs/structs/structwalk/walktype_bench_test.go b/libs/structs/structwalk/walktype_bench_test.go
index 5f4b05bf986..5947ba5ca47 100644
--- a/libs/structs/structwalk/walktype_bench_test.go
+++ b/libs/structs/structwalk/walktype_bench_test.go
@@ -21,8 +21,7 @@ func countFields(typ reflect.Type) (int, error) {
 func benchmarkWalkType(b *testing.B, tt reflect.Type) {
 	total := 0
 
-	b.ResetTimer()
-	for range b.N {
+	for b.Loop() {
 		count, err := countFields(tt)
 		if err != nil {
 			b.Fatalf("WalkType failed: %v", err)
@@ -30,7 +29,6 @@ func benchmarkWalkType(b *testing.B, tt reflect.Type) {
 		total += count
 	}
 
-	b.StopTimer()
 	// Root now correctly includes embedded struct fields, so it has many more fields than JobSettings
 	// (3,487 vs 533) because it contains JobSettings plus other resource types and config fields
 	b.Logf("Counted fields in %s: %d (%d/%d)", tt, total/b.N, total, b.N)