diff --git a/graphai/core/embedding/embedding.py b/graphai/core/embedding/embedding.py
index 5a9ab042..faa0b121 100644
--- a/graphai/core/embedding/embedding.py
+++ b/graphai/core/embedding/embedding.py
@@ -236,6 +236,7 @@ def _get_model_output(self, model, text):
     def _embed(self, model, text, force_split):
         text_too_large = False
         result = self._get_model_output(model, text)
+
         if result is None:
             if force_split:
                 model_max_tokens = self._get_model_max_tokens(model)
@@ -251,6 +252,12 @@ def _embed(self, model, text, force_split):
                     result = results.sum(axis=0).flatten()
             else:
                 text_too_large = True
+
+        # Normalise for all vectors to have unit length
+        norm = np.linalg.norm(result)
+        if norm > 0:
+            result = result / norm
+
         return result, text_too_large
 
     def embed(self, text, model_type='all-MiniLM-L12-v2', force_split=True):