diff --git a/src/common/transformations/include/transformations/symbolic_transformations/reshape_optimizations.hpp b/src/common/transformations/include/transformations/symbolic_transformations/reshape_optimizations.hpp
index 8cdcaee05e7..6ad409b4d96 100644
--- a/src/common/transformations/include/transformations/symbolic_transformations/reshape_optimizations.hpp
+++ b/src/common/transformations/include/transformations/symbolic_transformations/reshape_optimizations.hpp
@@ -16,7 +16,39 @@ class TRANSFORMATIONS_API ReshapeOptimizations;
 /**
  * @ingroup ov_transformation_common_api
  * @brief Searches for Flatten-like Reshape operations and simplifies 2nd input of such Reshape using special zero
- * feature
+ * feature.
+ * The transformation works in 2 cases:
+ * 1. all in/out dims are static, or we can match them via the symbols.
+   2. only one out dim doesn't have the corresponding input static dim,
+      and we can't match it using symbols. Besides that the output shape must not contain zero dims,
+      because then value -1 in 2nd input to Reshape op can't guarantee an unambiguous determination of the remaining dim
+ value.
+
+      for example:
+            Before:
+            +-------------+    +----------+
+            |data         |    | Concat   |
+            |shape: (0, 0)|    | shape (2)|<- the values might be determined in runtime
+            +-----+-------+    +-----+----+   the empty data tensor can be reshaped to
+                  |                  |        any other empty shape, e.g. (0, 800)
+            +-----v---------------+  |
+            | Reshape             |  |
+            | shape (0,-1)        <--+
+            | special zero = False|
+            +---------------------+
+
+            After:
+                               +---------------+
+           +--------------+    | Constant      |
+           | data         |    | shape (2)     |
+           | shape: (0, 0)|    | values (0, -1)|<- -1 means copy the corresponding input dim
+           +-----+--------+    +-------+-------+
+                 |                     |
+            +----v----------------+    |
+            | Reshape             <----+
+            | shape (0,0)         |  <- it might cause inconsistency
+            | special zero = True |
+            +---------------------+
  */
 class ov::pass::ReshapeOptimizations : public ov::pass::MatcherPass {
 public:
diff --git a/src/common/transformations/src/transformations/symbolic_transformations/reshape_optimizations.cpp b/src/common/transformations/src/transformations/symbolic_transformations/reshape_optimizations.cpp
index eee0c5d4db4..482f56367be 100644
--- a/src/common/transformations/src/transformations/symbolic_transformations/reshape_optimizations.cpp
+++ b/src/common/transformations/src/transformations/symbolic_transformations/reshape_optimizations.cpp
@@ -34,16 +34,23 @@ ov::pass::ReshapeOptimizations::ReshapeOptimizations() {
         const auto& out_shape = reshape->get_output_partial_shape(0);
         const auto& out_rank = out_shape.size();
 
+        int64_t cnt_static_zeros = 0;
         std::vector<int64_t> output_pattern(out_rank, -1);
         for (size_t i = 0; i < out_rank; ++i) {
-            if (out_shape[i].is_static())
+            if (out_shape[i].is_static()) {
                 output_pattern[i] = out_shape[i].get_length();
-            else if (i >= in_rank)
+                if (output_pattern[i] == 0) {
+                    ++cnt_static_zeros;
+                }
+            } else if (i >= in_rank) {
                 break;
-            else if (dims_are_equal(in_shape[i], out_shape[i]))
+            } else if (dims_are_equal(in_shape[i], out_shape[i])) {
                 output_pattern[i] = 0;
+            }
         }
-        if (std::count(output_pattern.begin(), output_pattern.end(), -1) <= 1) {
+
+        int64_t cnt_neg_ones = std::count(output_pattern.begin(), output_pattern.end(), -1);
+        if (cnt_neg_ones == 0 || (cnt_neg_ones == 1 && cnt_static_zeros == 0)) {
             auto new_pattern = ov::op::v0::Constant::create(element::i64, Shape{output_pattern.size()}, output_pattern);
             ov::copy_runtime_info(reshape->get_input_node_shared_ptr(1), new_pattern);
             reshape->set_special_zero(true);
diff --git a/src/common/transformations/tests/symbolic_transformations/reshape_optimizations.cpp b/src/common/transformations/tests/symbolic_transformations/reshape_optimizations.cpp
index 2f1fdba8dc2..8535abfe644 100644
--- a/src/common/transformations/tests/symbolic_transformations/reshape_optimizations.cpp
+++ b/src/common/transformations/tests/symbolic_transformations/reshape_optimizations.cpp
@@ -149,3 +149,26 @@ TEST_F(TransformationTestsF, NegativeTest) {
         manager.register_pass<pass::ReshapeOptimizations>();
     }
 }
+
+TEST_F(TransformationTestsF, ZeroDimsInOutputShape) {
+    // [A, B]
+    auto shape = PartialShape{0, 0};
+    {
+        auto data = make_shared<v0::Parameter>(element::f32, shape);
+        auto b = make_shared<v0::Parameter>(element::f32, PartialShape{-1});
+
+        auto a = ov::op::v0::Constant::create(element::i64, Shape{1}, {0});
+
+        auto shape_of = make_shared<v3::ShapeOf>(b);
+        auto indices = ov::op::v0::Constant::create(element::i64, {1}, {0});
+        auto axis = ov::op::v0::Constant::create(element::i64, {}, {0});
+        auto b_dim = make_shared<v1::Gather>(shape_of, indices, axis);
+
+        auto pattern = make_shared<v0::Concat>(OutputVector{a, b_dim}, 0);
+
+        auto reshape = make_shared<v1::Reshape>(data, pattern, false);
+
+        model = make_shared<Model>(NodeVector{reshape}, ParameterVector{data, b});
+        manager.register_pass<pass::ReshapeOptimizations>();
+    }
+}