diff --git a/token/services/storage/db/sql/common/tokens.go b/token/services/storage/db/sql/common/tokens.go
index cbce8c7e1c..fbffdaf664 100644
--- a/token/services/storage/db/sql/common/tokens.go
+++ b/token/services/storage/db/sql/common/tokens.go
@@ -190,7 +190,12 @@ func (db *TokenStore) UnspentTokensIterator(ctx context.Context) (tdriver.Unspen
 // append. UNION ALL is used (not UNION) to skip the per-row sort/hash
 // dedup pass; duplicates between the two branches (and within branch 1 when
 // a token has multiple ownership rows) are filtered at the iterator layer.
-func (db *TokenStore) UnspentTokensIteratorBy(ctx context.Context, walletID string, tokenType token.Type) (tdriver.UnspentTokensIterator, error) {
+// buildUnspentTokensIteratorByQuery builds the SQL query and args for
+// UnspentTokensIteratorBy without executing it. Extracted so benchmarks can
+// compare executing this exact query dynamically (query built and run fresh
+// each call, the production path) against running it via a statement
+// prepared once ahead of time, using identical SQL in both cases.
+func buildUnspentTokensIteratorByQuery(db *TokenStore, walletID string, tokenType token.Type) (string, []any) {
 	tokenTable := q.Table(db.table.Tokens)
 	ownershipTable := q.Table(db.table.Ownership)
 	joinCond := cond.And(
@@ -258,7 +263,12 @@ func (db *TokenStore) UnspentTokensIteratorBy(ctx context.Context, walletID stri
 	branch1.FormatTo(db.ci, sb)
 	sb.WriteString(" UNION ALL ")
 	branch2.FormatTo(db.ci, sb)
-	query, args := sb.Build()
+
+	return sb.Build()
+}
+
+func (db *TokenStore) UnspentTokensIteratorBy(ctx context.Context, walletID string, tokenType token.Type) (tdriver.UnspentTokensIterator, error) {
+	query, args := buildUnspentTokensIteratorByQuery(db, walletID, tokenType)
 
 	logging.Debug(logger, query, args)
 	rows, err := db.readDB.QueryContext(ctx, query, args...)
diff --git a/token/services/storage/db/sql/common/tokens_prepared_bench_util.go b/token/services/storage/db/sql/common/tokens_prepared_bench_util.go
new file mode 100644
index 0000000000..ec73824325
--- /dev/null
+++ b/token/services/storage/db/sql/common/tokens_prepared_bench_util.go
@@ -0,0 +1,99 @@
+/*
+Copyright IBM Corp. All Rights Reserved.
+
+SPDX-License-Identifier: Apache-2.0
+*/
+
+package common
+
+import (
+	"context"
+	"database/sql"
+	"testing"
+	"time"
+
+	"github.com/LFDT-Panurus/panurus/token/services/benchmark"
+	tokentype "github.com/LFDT-Panurus/panurus/token/token"
+)
+
+// RunUnspentTokensIteratorByPreparedComparison benchmarks UnspentTokensIteratorBy
+// against a version of the same query executed via a statement prepared once
+// outside the timed loop, using the same seed data, worker count, and duration
+// as RunTokenStoreBenchmarks so the numbers are directly comparable.
+//
+// This exists to answer: "how much would moving to prepared statements help
+// for our primary token-selection query?" (see #1183). The query built by
+// UnspentTokensIteratorBy is dynamic (its shape depends on whether walletID /
+// tokenType are empty), so this comparison fixes both parameters to non-empty
+// values matching the seeded data, which is the common case in production.
+func RunUnspentTokensIteratorByPreparedComparison(b *testing.B, store *TokenStore) {
+	b.Helper()
+
+	const walletID = "wallet0"
+	tokenType := tokentype.Type("GOLD")
+
+	b.Run("UnspentTokensIteratorBy_Dynamic", func(b *testing.B) {
+		SeedBenchTokens(b, store, 1000)
+		cfg := benchmark.NewConfig(4, 5*time.Second, 500*time.Millisecond)
+		result := benchmark.RunBenchmark(
+			cfg,
+			func() *TokenStore { return store },
+			func(s *TokenStore) error {
+				it, err := s.UnspentTokensIteratorBy(context.Background(), walletID, tokenType)
+				if err != nil {
+					return err
+				}
+				defer it.Close()
+				for {
+					tok, err := it.Next()
+					if err != nil {
+						return err
+					}
+					if tok == nil {
+						break
+					}
+				}
+				return nil
+			},
+		)
+		result.Print()
+	})
+
+	b.Run("UnspentTokensIteratorBy_Prepared", func(b *testing.B) {
+		SeedBenchTokens(b, store, 1000)
+
+		query, args := buildUnspentTokensIteratorByQuery(store, walletID, tokenType)
+
+		stmt, err := store.readDB.PrepareContext(context.Background(), query)
+		if err != nil {
+			b.Fatalf("failed preparing statement: %v", err)
+		}
+		defer func() { _ = stmt.Close() }()
+
+		cfg := benchmark.NewConfig(4, 5*time.Second, 500*time.Millisecond)
+		result := benchmark.RunBenchmark(
+			cfg,
+			func() *sql.Stmt { return stmt },
+			func(s *sql.Stmt) error {
+				rows, err := s.QueryContext(context.Background(), args...)
+				if err != nil {
+					return err
+				}
+				defer func() { _ = rows.Close() }()
+
+				it := &dedupedTokenRowsIterator{rows: rows, seen: make(map[string]struct{})}
+				for {
+					tok, err := it.Next()
+					if err != nil {
+						return err
+					}
+					if tok == nil {
+						break
+					}
+				}
+				return nil
+			},
+		)
+		result.Print()
+	})
+}
diff --git a/token/services/storage/db/sql/postgres/tokens_bench_test.go b/token/services/storage/db/sql/postgres/tokens_bench_test.go
index 6cfe281d81..ba359eaaa9 100644
--- a/token/services/storage/db/sql/postgres/tokens_bench_test.go
+++ b/token/services/storage/db/sql/postgres/tokens_bench_test.go
@@ -48,3 +48,9 @@ func BenchmarkTokenStore(b *testing.B) {
 	defer cleanup()
 	sqlcommon.RunTokenStoreBenchmarks(b, store)
 }
+
+func BenchmarkTokenStorePreparedComparison(b *testing.B) {
+	store, cleanup := openBenchTokenStore(b)
+	defer cleanup()
+	sqlcommon.RunUnspentTokensIteratorByPreparedComparison(b, store)
+}