diff --git a/src/plugins/intel_gpu/src/graph/layout_optimizer.cpp b/src/plugins/intel_gpu/src/graph/layout_optimizer.cpp
index d66440e67f3..6f862b13917 100644
--- a/src/plugins/intel_gpu/src/graph/layout_optimizer.cpp
+++ b/src/plugins/intel_gpu/src/graph/layout_optimizer.cpp
@@ -1746,7 +1746,7 @@ void layout_optimizer::select_preferred_formats_for_onednn(program_node& node, d
 
             // Conv or deconv gets a preferred format for its data input based on source memory description
             // But an input format for fused post-ops should be same with an output format of conv/deconv
-            size_t prim_input(0);
+            size_t prim_input(-1);
             if (node.is_type<convolution>())
                 prim_input = node.get_dependency_index(node.as<convolution>().input());
             if (node.is_type<deconvolution>())
@@ -1759,6 +1759,12 @@ void layout_optimizer::select_preferred_formats_for_onednn(program_node& node, d
             else  // Dep for fused post ops
                 src_fmt = onednn::find_data_format(prim_desc.dst_desc());
 
+            // WA: shallow convolution needs to set input format by bfyx.
+            //     onednn recommended byxf for input format. It will insert reorder before shallow conv.
+            if (node.is_type<convolution>() && node.get_input_layouts()[0].feature() == 3) {
+                src_fmt = format::get_default_format(node.get_input_layouts()[0].get_rank(), false, false);
+            }
+
             node.set_preferred_input_fmt(idx, src_fmt);
 
             auto dst_fmt = onednn::find_data_format(prim_desc.dst_desc());