diff --git a/tests/validation/fixtures/ConvolutionLayerFixture.h b/tests/validation/fixtures/ConvolutionLayerFixture.h
index 9dac137dcc..67f7d1337b 100644
--- a/tests/validation/fixtures/ConvolutionLayerFixture.h
+++ b/tests/validation/fixtures/ConvolutionLayerFixture.h
@@ -198,6 +198,11 @@ class ConvolutionValidationGenericFixture : public framework::Fixture
                bool                padded_weights               = false,
                bool                updated_sq_info_after_config = false)
     {
+#ifndef ARM_COMPUTE_CPU_ENABLED
+        ARM_COMPUTE_UNUSED(updated_sq_info_after_config);
+        ARM_COMPUTE_ERROR_ON(updated_sq_info_after_config);
+#endif // ARM_COMPUTE_CPU_ENABLED
+
         if (std::is_same<TensorType, Tensor>::value && // Cpu
             (data_type == DataType::F16 || weights_data_type == DataType::F16) && !CPUInfo::get().has_fp16())
         {
@@ -236,6 +241,7 @@ class ConvolutionValidationGenericFixture : public framework::Fixture
             _use_dynamic_output_quant = true;
         }
 
+#ifdef ARM_COMPUTE_CPU_ENABLED
         if (updated_sq_info_after_config)
         {
             _target = compute_gemmlowp_target_for_updated_sq_info_after_config(
@@ -243,6 +249,7 @@ class ConvolutionValidationGenericFixture : public framework::Fixture
                 pre_pad_layer, padded_weights);
         }
         else
+#endif // ARM_COMPUTE_CPU_ENABLED
         {
             _target = compute_target(input_shape, weights_shape, bias_shape, output_shape, info, reshape_weights,
                                      dilation, act_info, pre_pad_layer, padded_weights);
@@ -484,6 +491,7 @@ class ConvolutionValidationGenericFixture : public framework::Fixture
         return dst;
     }
 
+#ifdef ARM_COMPUTE_CPU_ENABLED
     // Compute the target when updating static quantization information after configuration.
     TensorType compute_gemmlowp_target_for_updated_sq_info_after_config(TensorShape               input_shape,
                                                                         TensorShape               weights_shape,
@@ -496,6 +504,9 @@ class ConvolutionValidationGenericFixture : public framework::Fixture
                                                                         PaddingList pre_pad_layer  = PaddingList({}),
                                                                         bool        padded_weights = false)
     {
+        // Runtime assert rather than static_assert because this template gets
+        // instantiated with FunctionType other than NEGEMMConvolutionLayer. The
+        // function call itself is prevented at runtime with conditions.
         ARM_COMPUTE_ASSERT((std::is_same<FunctionType, NEGEMMConvolutionLayer>::value == true));
         ARM_COMPUTE_ERROR_ON((input_shape[2] % weights_shape[2]) != 0);
 
@@ -612,6 +623,7 @@ class ConvolutionValidationGenericFixture : public framework::Fixture
         }
         return dst;
     }
+#endif // ARM_COMPUTE_CPU_ENABLED
 
     SimpleTensor<TO> compute_reference(const TensorShape        &input_shape,
                                        const TensorShape        &weights_shape,
diff --git a/tests/validation/fixtures/GEMMLowpFixture.h b/tests/validation/fixtures/GEMMLowpFixture.h
index 0a80252388..79671c58a4 100644
--- a/tests/validation/fixtures/GEMMLowpFixture.h
+++ b/tests/validation/fixtures/GEMMLowpFixture.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2025 Arm Limited.
+ * Copyright (c) 2017-2026 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -105,6 +105,7 @@ struct TensorFillInfo
     int32_t hash{0};
 };
 
+#ifdef ARM_COMPUTE_CPU_ENABLED
 template <typename TensorType,
           typename AccessorType,
           typename FunctionType,
@@ -129,6 +130,9 @@ TensorType compute_gemmlowp_target_for_updated_sq_info_after_config(
     DataType                   data_type_output            = DataType::UNKNOWN,
     const ActivationLayerInfo &act_info                    = ActivationLayerInfo())
 {
+    // Runtime assert rather than static_assert because this template gets
+    // instantiated with FunctionType other than NEGEMMLowpMatrixMultiplyCore. The
+    // function call itself is prevented at runtime with conditions.
     ARM_COMPUTE_ASSERT((std::is_same<FunctionType, NEGEMMLowpMatrixMultiplyCore>::value == true));
     ARM_COMPUTE_ASSERT(is_data_type_quantized_asymmetric(data_type_a));
     ARM_COMPUTE_ASSERT(is_data_type_quantized_asymmetric(data_type_b));
@@ -222,6 +226,7 @@ TensorType compute_gemmlowp_target_for_updated_sq_info_after_config(
     gemmlowp.run();
     return output;
 }
+#endif // ARM_COMPUTE_CPU_ENABLED
 
 template <typename TensorType,
           typename AccessorType,
@@ -730,6 +735,12 @@ class GEMMLowpGenericMatrixMultiplyCoreFusedOffsetOutputValidationFixture : publ
                               bool                           updated_sq_info_after_config = false,
                               const ActivationLayerInfo     &act_info                     = ActivationLayerInfo())
     {
+#ifndef ARM_COMPUTE_CPU_ENABLED
+        ARM_COMPUTE_UNUSED(updated_sq_info_after_config, act_info);
+        ARM_COMPUTE_ERROR_ON(updated_sq_info_after_config);
+#endif // ARM_COMPUTE_CPU_ENABLED
+
+#ifdef ARM_COMPUTE_CPU_ENABLED
         if (updated_sq_info_after_config)
         {
             return compute_gemmlowp_target_for_updated_sq_info_after_config<
@@ -739,6 +750,7 @@ class GEMMLowpGenericMatrixMultiplyCoreFusedOffsetOutputValidationFixture : publ
                                  arm_compute::DataType::UNKNOWN, act_info);
         }
         else
+#endif // ARM_COMPUTE_CPU_ENABLED
         {
             return compute_gemmlowp_target<TensorType, AccessorType, FunctionType, reinterpret_input_as_3d,
                                            reinterpret_output_as_3d, qasymm8_t, true, run_twice>(