diff --git a/onnxruntime/contrib_ops/webgpu/bert/flash_attention.cc b/onnxruntime/contrib_ops/webgpu/bert/flash_attention.cc
index 606dbfde15c2c..3f5b28f92f55d 100644
--- a/onnxruntime/contrib_ops/webgpu/bert/flash_attention.cc
+++ b/onnxruntime/contrib_ops/webgpu/bert/flash_attention.cc
@@ -421,7 +421,7 @@ Status ApplyFlashAttention(const Tensor* Q, const Tensor* K, const Tensor* V, co
                                                                       indirect_buffer_ptr, tile_size));
     Q = &query_output;
   } else {
-    ORT_RETURN_IF_ERROR(CopyKVCache(context, parameters, K, past_key, present_key, V, past_value, present_value, tile_size, use_indirect_dispatch ? seqlen_k : nullptr, indirect_buffer_ptr));
+    ORT_RETURN_IF_ERROR(CopyKVCache(context, parameters, K, past_key, present_key, V, past_value, present_value, tile_size, use_seqlen_k ? seqlen_k : nullptr, indirect_buffer_ptr));
   }
 
   if (parameters.sequence_length_ > 1) {