diff --git a/backend/Generator/llm_generator.py b/backend/Generator/llm_generator.py
index e6333030..a363ccb3 100644
--- a/backend/Generator/llm_generator.py
+++ b/backend/Generator/llm_generator.py
@@ -2,7 +2,7 @@
 import re
 import threading
 from llama_cpp import Llama
-
+from .output_validator import validate_mcq, validate_shortq, validate_boolq
 
 class LLMQuestionGenerator:
     """Generates various types of questions using Qwen3-0.6B via llama.cpp.
@@ -73,7 +73,8 @@ def generate_short_questions(self, input_text, max_questions=4):
                 return []
 
             raw = choices[0].get("message", {}).get("content", "")
-            return self._parse_response(raw, max_questions)
+            parsed_questions = self._parse_response(raw, max_questions)
+            return validate_shortq(parsed_questions)
 
         except (AttributeError, TypeError, ValueError):
             return []
@@ -113,7 +114,8 @@ def generate_mcq_questions(self, input_text, max_questions=4):
                 return []
 
             raw = choices[0].get("message", {}).get("content", "")
-            return self._parse_mcq_response(raw, max_questions)
+            parsed_questions = self._parse_mcq_response(raw, max_questions)
+            return validate_mcq(parsed_questions)
 
         except (AttributeError, TypeError, ValueError):
             return []
@@ -152,7 +154,8 @@ def generate_boolean_questions(self, input_text, max_questions=4):
                 return []
 
             raw = choices[0].get("message", {}).get("content", "")
-            return self._parse_bool_response(raw, max_questions)
+            parsed_questions = self._parse_bool_response(raw, max_questions)
+            return validate_boolq(parsed_questions)
 
         except (AttributeError, TypeError, ValueError):
             return []
diff --git a/backend/Generator/output_validator.py b/backend/Generator/output_validator.py
new file mode 100644
index 00000000..96d1f52a
--- /dev/null
+++ b/backend/Generator/output_validator.py
@@ -0,0 +1,131 @@
+"""
+Output validation for LLM-generated questions.
+Ensures that only well-formed, complete questions are returned to API consumers.
+"""
+
+
+def validate_mcq(questions):
+    """
+    Validate multiple-choice questions.
+    
+    Args:
+        questions: List of MCQ question dictionaries
+        
+    Returns:
+        List of valid MCQ questions (filtered)
+        
+    Validation Rules:
+        - question field must exist and be non-empty string
+        - options field must exist and be a list with at least 2 elements
+        - correct_answer field must exist and be non-empty string
+    """
+    if not questions:
+        return []
+    
+    valid_questions = []
+    
+    for q in questions:
+        if not isinstance(q, dict):
+            continue
+            
+        # Check question field
+        question = q.get("question")
+        if not question or not isinstance(question, str) or not question.strip():
+            continue
+            
+        # Check options field
+        options = q.get("options")
+        if not options or not isinstance(options, list) or len(options) < 2:
+            continue
+        
+        # Verify all options are non-empty strings
+        if not all(isinstance(opt, str) and opt.strip() for opt in options):
+            continue
+
+        # Check correct_answer field
+        correct_answer = q.get("correct_answer")
+        if not correct_answer or not isinstance(correct_answer, str) or not correct_answer.strip():
+            continue
+            
+        # All checks passed - this is a valid question
+        valid_questions.append(q)
+    
+    return valid_questions
+
+
+def validate_shortq(questions):
+    """
+    Validate short answer questions.
+    
+    Args:
+        questions: List of short answer question dictionaries
+        
+    Returns:
+        List of valid short answer questions (filtered)
+        
+    Validation Rules:
+        - question field must exist and be non-empty string
+        - answer field must exist and be non-empty string
+    """
+    if not questions:
+        return []
+    
+    valid_questions = []
+    
+    for q in questions:
+        if not isinstance(q, dict):
+            continue
+            
+        # Check question field
+        question = q.get("question")
+        if not question or not isinstance(question, str) or not question.strip():
+            continue
+            
+        # Check answer field
+        answer = q.get("answer")
+        if not answer or not isinstance(answer, str) or not answer.strip():
+            continue
+            
+        # All checks passed - this is a valid question
+        valid_questions.append(q)
+    
+    return valid_questions
+
+
+def validate_boolq(questions):
+    """
+    Validate boolean (true/false) questions.
+    
+    Args:
+        questions: List of boolean question dictionaries
+        
+    Returns:
+        List of valid boolean questions (filtered)
+        
+    Validation Rules:
+        - question field must exist and be non-empty string
+        - answer field must exist and be boolean type (True or False)
+    """
+    if not questions:
+        return []
+    
+    valid_questions = []
+    
+    for q in questions:
+        if not isinstance(q, dict):
+            continue
+            
+        # Check question field
+        question = q.get("question")
+        if not question or not isinstance(question, str) or not question.strip():
+            continue
+            
+        # Check answer field - must be boolean type, not string or int
+        answer = q.get("answer")
+        if not isinstance(answer, bool):
+            continue
+            
+        # All checks passed - this is a valid question
+        valid_questions.append(q)
+    
+    return valid_questions