diff --git a/buster/retriever/service.py b/buster/retriever/service.py
index 5e1ce5d..362571f 100644
--- a/buster/retriever/service.py
+++ b/buster/retriever/service.py
@@ -92,4 +92,7 @@ def get_topk_documents(self, query: str, source: str, top_k: int) -> pd.DataFram
         matched_documents["similarity"] = matched_documents["_id"].apply(lambda x: matching_scores[str(x)])
         matched_documents["embedding"] = matched_documents["_id"].apply(lambda x: matching_embeddings[str(x)])
 
+        # sort by similarity
+        matched_documents = matched_documents.sort_values(by="similarity", ascending=False, ignore_index=True)
+
         return matched_documents