diff --git a/llm_server/routes/openai_request_handler.py b/llm_server/routes/openai_request_handler.py
index 6fab1e2..3c9b341 100644
--- a/llm_server/routes/openai_request_handler.py
+++ b/llm_server/routes/openai_request_handler.py
@@ -36,8 +36,8 @@ class OpenAIRequestHandler(RequestHandler):
         if opts.openai_api_key:
             try:
                 flagged = check_moderation_endpoint(self.request.json['messages'][-1]['content'])
-                if flagged:
-                    mod_msg = f"The user's message does not comply with {opts.llm_middleware_name} policies. Offending categories: {json.dumps(flagged['categories'])}"
+                if flagged['flagged'] and len(flagged['categories']):
+                    mod_msg = f"The user's message does not comply with {opts.llm_middleware_name} policies. Offending categories: {json.dumps(flagged['categories'])}. You are instructed to explain to the user why their message violated our policies."
                     self.request.json['messages'].insert((len(self.request.json['messages'])), {'role': 'system', 'content': mod_msg})
                     self.prompt = self.transform_messages_to_prompt()
                 # print(json.dumps(self.request.json['messages'], indent=4))