diff --git a/src/plugins/intel_gpu/src/kernel_selector/kernels/fully_connected/fully_connected_kernel_bf_tiled.cpp b/src/plugins/intel_gpu/src/kernel_selector/kernels/fully_connected/fully_connected_kernel_bf_tiled.cpp
index b26b11ce97df6a..be41f5faef3272 100644
--- a/src/plugins/intel_gpu/src/kernel_selector/kernels/fully_connected/fully_connected_kernel_bf_tiled.cpp
+++ b/src/plugins/intel_gpu/src/kernel_selector/kernels/fully_connected/fully_connected_kernel_bf_tiled.cpp
@@ -293,8 +293,17 @@ bool TuneParamsSelector::VerifyTuneParams(const fully_connected_params& params,
     size_t output_f = bf_size.second;
 
     auto batch_size = params.is_shape_agnostic ? Align(output_b, tparams.tile_b) : output_b;
-    if (batch_size % (tparams.tile_b * tparams.dispatch_bsv) != 0)
-        return false;
+    // If batch size is prime number, still can apply tile execution to avoid poor performance.
+    if (batch_size % (tparams.tile_b * tparams.dispatch_bsv) != 0) {
+        if ((tparams.dispatch_bsv != 1) || batch_size == 1)
+            return false;
+        size_t tile = simd;
+        while (batch_size % tile != 0)
+            tile--;
+        if (tile > 1)
+            return false;
+    }
+
     if (CeilDiv(output_f, tparams.tile_ofm * simd) % tparams.dispatch_fsv != 0)
         return false;
 
diff --git a/src/plugins/intel_gpu/tests/unit/test_cases/fully_connected_gpu_test.cpp b/src/plugins/intel_gpu/tests/unit/test_cases/fully_connected_gpu_test.cpp
index dde1b6215148b3..eed9760348f669 100644
--- a/src/plugins/intel_gpu/tests/unit/test_cases/fully_connected_gpu_test.cpp
+++ b/src/plugins/intel_gpu/tests/unit/test_cases/fully_connected_gpu_test.cpp
@@ -1227,6 +1227,59 @@ TEST(fully_connected_gpu, bf_tiled_with_pad) {
     }
 }
 
+TEST(fully_connected_gpu, bf_tiled_with_unaligned_batch) {
+    tests::random_generator rg(GET_SUITE_NAME);
+    auto& engine = get_test_engine();
+    // Test parameters with unaligned batch size.
+    const int batch_num = 17;
+    const int feature_num = 1;
+    const int input_x = 1;
+    const int input_y = 64;
+    const int output_y = input_y;
+
+    // Allocate memory
+    auto input_mem =
+        engine.allocate_memory({{batch_num, feature_num, input_y, input_x}, data_types::f16, format::bfyx});
+    auto weights_mem = engine.allocate_memory({{output_y, input_y}, data_types::f16, format::bfyx});
+
+    // Generate random input data and set values
+    auto input_data = rg.generate_random_4d<ov::float16>(batch_num, feature_num, input_y, input_x, -1, 1);
+    auto weights_data = rg.generate_random_4d<ov::float16>(output_y, input_y, 1, 1, -1, 1);
+
+    auto input_data_bfyx = flatten_4d(format::bfyx, input_data);
+    auto weights_data_bfyx = flatten_4d(format::bfyx, weights_data);
+
+    std::vector<ov::float16> empty_bias(output_y, 0);
+    set_values(input_mem, input_data_bfyx);
+    set_values(weights_mem, weights_data_bfyx);
+    auto reference_output = dynamic_fully_connected_reference_calc<ov::float16>(batch_num * feature_num,
+                                                                                input_y,
+                                                                                output_y,
+                                                                                input_data_bfyx,
+                                                                                weights_data_bfyx,
+                                                                                empty_bias);
+    topology topology(input_layout("input", input_mem->get_layout()),
+                      data("weights", weights_mem),
+                      fully_connected("fc_prim", input_info("input"), "weights", "", 3, 3));
+
+    // Set data optimization to allow weights reordering to optimal format
+    ExecutionConfig config = get_test_default_config(engine);
+    config.set_property(ov::intel_gpu::optimize_data(true));
+    ov::intel_gpu::ImplementationDesc fc_impl_desc = {format::bfyx, "fully_connected_gpu_bf_tiled", impl_types::ocl};
+    config.set_property(ov::intel_gpu::force_implementations(ov::intel_gpu::ImplForcingMap{{"fc_prim", fc_impl_desc}}));
+
+    network network(engine, topology, config);
+    network.set_input_data("input", input_mem);
+
+    auto outputs = network.execute();
+    auto output_mem = outputs.at("fc_prim").get_memory();
+    cldnn::mem_lock<ov::float16> output_ptr(output_mem, get_test_stream());
+    ASSERT_EQ(output_mem->count(), batch_num * feature_num * output_y);
+
+    for (size_t i = 0; i < batch_num * feature_num * output_y; ++i) {
+        ASSERT_FLOAT_EQ(reference_output[i], output_ptr[i]) << " i = " << i;
+    }
+}
 
 TEST(fully_connected_gpu, DISABLED_fs_byx_fsv32_b34)
 {