diff --git a/libcudacxx/include/cuda/__memory_pool/device_memory_pool.h b/libcudacxx/include/cuda/__memory_pool/device_memory_pool.h
index d29b63c0f19..d4f7e34e142 100644
--- a/libcudacxx/include/cuda/__memory_pool/device_memory_pool.h
+++ b/libcudacxx/include/cuda/__memory_pool/device_memory_pool.h
@@ -137,10 +137,10 @@ struct device_memory_pool : device_memory_pool_ref
   }
 
   //! @brief Returns a \c device_memory_pool_ref for this \c device_memory_pool.
-  //! The result is the same as if this object was cast to a \c device_memory_pool_ref.
-  [[nodiscard]] _CCCL_HOST_API device_memory_pool_ref as_ref() noexcept
+  //! We return by reference to ensure that we can subsequently convert to a resource_ref
+  [[nodiscard]] _CCCL_HOST_API device_memory_pool_ref& as_ref() noexcept
   {
-    return device_memory_pool_ref(__pool_);
+    return static_cast<device_memory_pool_ref&>(*this);
   }
 
   device_memory_pool(const device_memory_pool&)            = delete;
diff --git a/libcudacxx/include/cuda/__memory_pool/managed_memory_pool.h b/libcudacxx/include/cuda/__memory_pool/managed_memory_pool.h
index 56d6f320c9b..8790faef7f0 100644
--- a/libcudacxx/include/cuda/__memory_pool/managed_memory_pool.h
+++ b/libcudacxx/include/cuda/__memory_pool/managed_memory_pool.h
@@ -131,10 +131,10 @@ struct managed_memory_pool : managed_memory_pool_ref
   }
 
   //! @brief Returns a \c managed_memory_pool_ref for this \c managed_memory_pool.
-  //! The result is the same as if this object was cast to a \c managed_memory_pool_ref.
-  [[nodiscard]] _CCCL_HOST_API managed_memory_pool_ref as_ref() noexcept
+  //! We return by reference to ensure that we can subsequently convert to a resource_ref
+  [[nodiscard]] _CCCL_HOST_API managed_memory_pool_ref& as_ref() noexcept
   {
-    return managed_memory_pool_ref(__pool_);
+    return static_cast<managed_memory_pool_ref&>(*this);
   }
 
   managed_memory_pool(const managed_memory_pool&)            = delete;
diff --git a/libcudacxx/include/cuda/__memory_pool/pinned_memory_pool.h b/libcudacxx/include/cuda/__memory_pool/pinned_memory_pool.h
index e1a33f55f8f..caf0fecf5ce 100644
--- a/libcudacxx/include/cuda/__memory_pool/pinned_memory_pool.h
+++ b/libcudacxx/include/cuda/__memory_pool/pinned_memory_pool.h
@@ -168,10 +168,10 @@ struct pinned_memory_pool : pinned_memory_pool_ref
   }
 
   //! @brief Returns a \c pinned_memory_pool_ref for this \c pinned_memory_pool.
-  //! The result is the same as if this object was cast to a \c pinned_memory_pool_ref.
-  _CCCL_HOST_API pinned_memory_pool_ref as_ref() noexcept
+  //! We return by reference to ensure that we can subsequently convert to a resource_ref
+  _CCCL_HOST_API pinned_memory_pool_ref& as_ref() noexcept
   {
-    return pinned_memory_pool_ref(__pool_);
+    return static_cast<pinned_memory_pool_ref&>(*this);
   }
 
   pinned_memory_pool(const pinned_memory_pool&)            = delete;
diff --git a/libcudacxx/test/libcudacxx/cuda/memory_resource/resources/device_memory_resource.cu b/libcudacxx/test/libcudacxx/cuda/memory_resource/resources/device_memory_resource.cu
index 1cfd2727d81..08380d925d9 100644
--- a/libcudacxx/test/libcudacxx/cuda/memory_resource/resources/device_memory_resource.cu
+++ b/libcudacxx/test/libcudacxx/cuda/memory_resource/resources/device_memory_resource.cu
@@ -65,7 +65,7 @@ static bool ensure_export_handle(::cudaMemPool_t pool, const ::cudaMemAllocation
   return allocation_handle == ::cudaMemHandleTypeNone ? status == ::cudaErrorInvalidValue : status == ::cudaSuccess;
 }
 
-C2H_CCCLRT_TEST("device_memory_resource construction", "[memory_resource]")
+C2H_CCCLRT_TEST("device_memory_pool construction", "[memory_resource]")
 {
   int current_device = 0;
   cuda::__ensure_current_context guard{cuda::device_ref{current_device}};
@@ -217,7 +217,7 @@ static void ensure_device_ptr(void* ptr)
   CHECK(attributes.type == cudaMemoryTypeDevice);
 }
 
-C2H_CCCLRT_TEST("device_memory_resource allocation", "[memory_resource]")
+C2H_CCCLRT_TEST("device_memory_pool allocation", "[memory_resource]")
 {
   cudaStream_t raw_stream;
   {
@@ -332,7 +332,7 @@ C2H_CCCLRT_TEST("device_memory_resource allocation", "[memory_resource]")
   }
 }
 
-C2H_CCCLRT_TEST("device_memory_resource comparison", "[memory_resource]")
+C2H_CCCLRT_TEST("device_memory_pool comparison", "[memory_resource]")
 {
   int current_device = 0;
   cuda::__ensure_current_context guard{cuda::device_ref{current_device}};
diff --git a/libcudacxx/test/libcudacxx/cuda/memory_resource/resources/memory_pools.cu b/libcudacxx/test/libcudacxx/cuda/memory_resource/resources/memory_pools.cu
index ad15960e690..7b31ec02cb4 100644
--- a/libcudacxx/test/libcudacxx/cuda/memory_resource/resources/memory_pools.cu
+++ b/libcudacxx/test/libcudacxx/cuda/memory_resource/resources/memory_pools.cu
@@ -670,3 +670,84 @@ C2H_CCCLRT_TEST("pinned_memory_pool with allocation handle", "[memory_resource]"
 
 // managed memory pool does not support allocation handles yet.
 #endif // !_CCCL_OS(WINDOWS)
+
+C2H_CCCLRT_TEST("device_memory_pool conversion to resource_ref", "[memory_resource]")
+{
+  int current_device = 0;
+  cuda::__ensure_current_context guard{cuda::device_ref{current_device}};
+
+  cuda::device_memory_pool pool{cuda::device_ref{0}};
+  cuda::mr::resource_ref<cuda::mr::device_accessible> ref1 = pool.as_ref();
+
+  cuda::device_memory_pool_ref pool_ref                    = pool.as_ref();
+  cuda::mr::resource_ref<cuda::mr::device_accessible> ref2 = pool_ref;
+  CHECK((ref1 == ref2));
+}
+
+#if _CCCL_CTK_AT_LEAST(13, 0) && !_CCCL_OS(WINDOWS)
+C2H_CCCLRT_TEST("managed_memory_pool conversion to resource_ref", "[memory_resource]")
+{
+  int current_device = 0;
+  cuda::__ensure_current_context guard{cuda::device_ref{current_device}};
+
+  cuda::managed_memory_pool pool{};
+
+  { // host device accessible
+    cuda::mr::resource_ref<cuda::mr::host_accessible, cuda::mr::device_accessible> ref1 = pool.as_ref();
+
+    cuda::managed_memory_pool_ref pool_ref                                              = pool.as_ref();
+    cuda::mr::resource_ref<cuda::mr::host_accessible, cuda::mr::device_accessible> ref2 = pool_ref;
+    CHECK((ref1 == ref2));
+  }
+
+  { // host  accessible
+    cuda::mr::resource_ref<cuda::mr::host_accessible> ref1 = pool.as_ref();
+
+    cuda::managed_memory_pool_ref pool_ref                 = pool.as_ref();
+    cuda::mr::resource_ref<cuda::mr::host_accessible> ref2 = pool_ref;
+    CHECK((ref1 == ref2));
+  }
+
+  { // device accessible
+    cuda::mr::resource_ref<cuda::mr::device_accessible> ref1 = pool.as_ref();
+
+    cuda::managed_memory_pool_ref pool_ref                   = pool.as_ref();
+    cuda::mr::resource_ref<cuda::mr::device_accessible> ref2 = pool_ref;
+    CHECK((ref1 == ref2));
+  }
+}
+#endif // _CCCL_CTK_AT_LEAST(13, 0) && !_CCCL_OS(WINDOWS)
+
+#if _CCCL_CTK_AT_LEAST(12, 6)
+C2H_CCCLRT_TEST("pinned_memory_pool conversion to resource_ref", "[memory_resource]")
+{
+  int current_device = 0;
+  cuda::__ensure_current_context guard{cuda::device_ref{current_device}};
+
+  cuda::pinned_memory_pool pool{0};
+
+  { // host device accessible
+    cuda::mr::resource_ref<cuda::mr::host_accessible, cuda::mr::device_accessible> ref1 = pool.as_ref();
+
+    cuda::pinned_memory_pool_ref pool_ref                                               = pool.as_ref();
+    cuda::mr::resource_ref<cuda::mr::host_accessible, cuda::mr::device_accessible> ref2 = pool_ref;
+    CHECK((ref1 == ref2));
+  }
+
+  { // host  accessible
+    cuda::mr::resource_ref<cuda::mr::host_accessible> ref1 = pool.as_ref();
+
+    cuda::pinned_memory_pool_ref pool_ref                  = pool.as_ref();
+    cuda::mr::resource_ref<cuda::mr::host_accessible> ref2 = pool_ref;
+    CHECK((ref1 == ref2));
+  }
+
+  { // device accessible
+    cuda::mr::resource_ref<cuda::mr::device_accessible> ref1 = pool.as_ref();
+
+    cuda::pinned_memory_pool_ref pool_ref                    = pool.as_ref();
+    cuda::mr::resource_ref<cuda::mr::device_accessible> ref2 = pool_ref;
+    CHECK((ref1 == ref2));
+  }
+}
+#endif // _CCCL_CTK_AT_LEAST(12, 6)