diff --git a/RELEASE.md b/RELEASE.md
index c287a896a8..fae46a37ed 100644
--- a/RELEASE.md
+++ b/RELEASE.md
@@ -16,6 +16,7 @@ Please follow the established format:
 - Fix incorrect rendering of datasets in modular pipelines. (#1439)
 - Fix broken SVG/PNG exports in light theme. (#1463)
 - Fix dataset and global toolbar error with standalone React component (#1351)
+- Fix `ImportError` as kedro-datasets is now lazily loaded (#1481).
 
 # Release 6.3.4
 
diff --git a/package/kedro_viz/integrations/kedro/data_loader.py b/package/kedro_viz/integrations/kedro/data_loader.py
index 4b78b44d95..ca4b1565b6 100644
--- a/package/kedro_viz/integrations/kedro/data_loader.py
+++ b/package/kedro_viz/integrations/kedro/data_loader.py
@@ -21,7 +21,7 @@
         plotly,
         tracking,
     )
-except ImportError:
+except ImportError:  # kedro_datasets is not installed.
     from kedro.extras.datasets import (  # Safe since ImportErrors are suppressed within kedro.
         json as json_dataset,
         matplotlib,
@@ -158,12 +158,12 @@ def load_data(
         return context.catalog, context.pipelines, session_store, stats_dict
 
 
-# The dataset type is available as an attribute if and only if the import from kedro
-# did not suppress an ImportError. i.e. hasattr(matplotlib, "MatplotlibWriter") is True
-# when matplotlib dependencies are installed.
+# Try to access the attribute to trigger the import of dependencies, only modify the _load
+# if dependencies are installed.
 # These datasets do not have _load methods defined (tracking and matplotlib) or do not
 # load to json (plotly), hence the need to define _load here.
-if hasattr(matplotlib, "MatplotlibWriter"):
+try:
+    getattr(matplotlib, "MatplotlibWriter")  # Trigger the lazy import
 
     def matplotlib_writer_load(dataset: matplotlib.MatplotlibWriter) -> str:
         load_path = get_filepath_str(dataset._get_load_path(), dataset._protocol)
@@ -172,15 +172,29 @@ def matplotlib_writer_load(dataset: matplotlib.MatplotlibWriter) -> str:
         return base64_bytes.decode("utf-8")
 
     matplotlib.MatplotlibWriter._load = matplotlib_writer_load
+except (ImportError, AttributeError):
+    pass
 
-if hasattr(plotly, "JSONDataSet"):
+try:
+    getattr(plotly, "JSONDataSet")  # Trigger import
     plotly.JSONDataSet._load = json_dataset.JSONDataSet._load
+except (ImportError, AttributeError):
+    pass
 
-if hasattr(plotly, "PlotlyDataSet"):
+try:
+    getattr(plotly, "PlotlyDataSet")  # Trigger import
     plotly.PlotlyDataSet._load = json_dataset.JSONDataSet._load
+except (ImportError, AttributeError):
+    pass
 
-if hasattr(tracking, "JSONDataSet"):
+try:
+    getattr(tracking, "JSONDataSet")  # Trigger import
     tracking.JSONDataSet._load = json_dataset.JSONDataSet._load
+except (ImportError, AttributeError):
+    pass
 
-if hasattr(tracking, "MetricsDataSet"):
+try:
+    getattr(tracking, "MetricsDataSet")  # Trigger import
     tracking.MetricsDataSet._load = json_dataset.JSONDataSet._load
+except (ImportError, AttributeError):
+    pass
diff --git a/package/kedro_viz/models/flowchart.py b/package/kedro_viz/models/flowchart.py
index bcf1d9b232..a2638851a1 100644
--- a/package/kedro_viz/models/flowchart.py
+++ b/package/kedro_viz/models/flowchart.py
@@ -595,7 +595,9 @@ def __post_init__(self, data_node: DataNode, dataset_stats: Dict):
             self.tracking_data = dataset.load()
         elif data_node.is_preview_node():
             try:
-                self.preview = dataset._preview(**data_node.get_preview_args())
+                if hasattr(dataset, "_preview"):
+                    self.preview = dataset._preview(**data_node.get_preview_args())
+
             except Exception as exc:  # pylint: disable=broad-except # pragma: no cover
                 logger.warning(
                     "'%s' could not be previewed. Full exception: %s: %s",