diff --git a/lettucedetect/detectors/base.py b/lettucedetect/detectors/base.py
index 2fee358..acb505c 100644
--- a/lettucedetect/detectors/base.py
+++ b/lettucedetect/detectors/base.py
@@ -71,6 +71,20 @@ def _validate_min_confidence(min_confidence: float) -> None:
         if not 0.0 <= min_confidence <= 1.0:
             raise ValueError(f"min_confidence must be in [0, 1], got {min_confidence}")
 
+    @staticmethod
+    def _validate_batch_lengths(prompts: list[str], answers: list[str]) -> None:
+        """Validate that ``prompts`` and ``answers`` have the same length.
+
+        :param prompts: The prompt strings of the batch.
+        :param answers: The answer strings of the batch.
+        :raises ValueError: If ``len(prompts) != len(answers)``.
+        """
+        if len(prompts) != len(answers):
+            raise ValueError(
+                "Number of prompts must match number of answers, "
+                f"got {len(prompts)} prompts and {len(answers)} answers"
+            )
+
     @staticmethod
     def _filter_spans_by_confidence(
         result: list, output_format: str, min_confidence: float
diff --git a/lettucedetect/detectors/llm.py b/lettucedetect/detectors/llm.py
index ef560b4..c83cc4f 100644
--- a/lettucedetect/detectors/llm.py
+++ b/lettucedetect/detectors/llm.py
@@ -603,8 +603,7 @@ def predict_prompt_batch(
             ``output_format="tokens"``.
         :raises ValueError: If ``len(prompts) != len(answers)``.
         """
-        if len(prompts) != len(answers):
-            raise ValueError("Number of prompts must match number of answers")
+        self._validate_batch_lengths(prompts, answers)
         if output_format not in ["tokens", "spans"]:
             raise ValueError(
                 f"LLMDetector doesn't support '{output_format}' format. Use 'tokens' or 'spans'"
diff --git a/lettucedetect/detectors/rag_fact_checker.py b/lettucedetect/detectors/rag_fact_checker.py
index c5b87d0..f22ac8b 100644
--- a/lettucedetect/detectors/rag_fact_checker.py
+++ b/lettucedetect/detectors/rag_fact_checker.py
@@ -107,8 +107,7 @@ def predict_prompt_batch(
         min_confidence: float = 0.0,
     ) -> list[list[dict[str, Any]]]:
         """Batch prediction using RAGFactChecker's batch processing."""
-        if len(prompts) != len(answers):
-            raise ValueError("Number of prompts must match number of answers")
+        self._validate_batch_lengths(prompts, answers)
         self._validate_min_confidence(min_confidence)
 
         contexts = [[prompt] for prompt in prompts]  # Convert prompts to context lists
@@ -116,7 +115,7 @@ def predict_prompt_batch(
 
         # Convert each result to lettuceDetect format
         converted_results = []
-        for i, (answer, rag_result) in enumerate(zip(answers, rag_results)):
+        for i, (answer, rag_result) in enumerate(zip(answers, rag_results, strict=True)):
             if output_format == "tokens":
                 converted = self._convert_to_tokens(answer, rag_result)
             elif output_format == "spans":
diff --git a/lettucedetect/detectors/transformer.py b/lettucedetect/detectors/transformer.py
index c488ddf..9a0e1d6 100644
--- a/lettucedetect/detectors/transformer.py
+++ b/lettucedetect/detectors/transformer.py
@@ -546,8 +546,7 @@ def predict_prompt_batch(
         :returns: List of prediction lists, one per input pair.
         :raises ValueError: If ``len(prompts) != len(answers)``.
         """
-        if len(prompts) != len(answers):
-            raise ValueError("Number of prompts must match number of answers")
+        self._validate_batch_lengths(prompts, answers)
         if not prompts:
             return []
         if output_format not in ("tokens", "spans"):
diff --git a/tests/test_inference_pytest.py b/tests/test_inference_pytest.py
index f16eb6e..c711f52 100644
--- a/tests/test_inference_pytest.py
+++ b/tests/test_inference_pytest.py
@@ -454,6 +454,23 @@ def test_filter_keeps_spans_without_confidence(self):
         assert BaseDetector._filter_spans_by_confidence(spans, "spans", 0.9) == spans
 
 
+class TestBatchLengthValidation:
+    """Unit tests for the shared BaseDetector._validate_batch_lengths helper."""
+
+    def test_mismatch_message_shows_both_lengths(self):
+        """A length mismatch raises a ValueError that shows the two lengths."""
+        with pytest.raises(ValueError, match="got 2 prompts and 1 answers"):
+            BaseDetector._validate_batch_lengths(["p1", "p2"], ["a1"])
+
+    @pytest.mark.parametrize(
+        ("prompts", "answers"),
+        [([], []), (["p1"], ["a1"]), (["p1", "p2"], ["a1", "a2"])],
+    )
+    def test_equal_lengths_raise_no_error(self, prompts, answers):
+        """The helper accepts equal lengths, including two empty lists."""
+        BaseDetector._validate_batch_lengths(prompts, answers)
+
+
 class TestTransformerMinConfidence:
     """min_confidence behaviour wired through TransformerDetector."""
 
diff --git a/tests/test_rag_fact_checker_pytest.py b/tests/test_rag_fact_checker_pytest.py
new file mode 100644
index 0000000..1513b3a
--- /dev/null
+++ b/tests/test_rag_fact_checker_pytest.py
@@ -0,0 +1,65 @@
+"""Pytest tests for RAGFactCheckerDetector batch prediction."""
+
+from unittest.mock import patch
+
+import pytest
+
+from lettucedetect.detectors.rag_fact_checker import RAGFactCheckerDetector
+
+HALLUCINATED_RESULT = {"hallucinated_triplets": [["Paris", "is capital of", "Germany"]]}
+CLEAN_RESULT = {"hallucinated_triplets": []}
+
+
+@pytest.fixture
+def detector():
+    """RAGFactCheckerDetector with a mock backend. The tests need no API key and no network."""
+    with patch("lettucedetect.ragfactchecker.RAGFactChecker"):
+        return RAGFactCheckerDetector()
+
+
+class TestPredictPromptBatchRAG:
+    """predict_prompt_batch() keeps one result for each (prompt, answer) pair."""
+
+    @pytest.mark.parametrize(
+        ("prompts", "answers"),
+        [
+            (["p1", "p2"], ["a1"]),
+            (["p1"], ["a1", "a2"]),
+            (["p1"], []),
+            ([], ["a1"]),
+        ],
+    )
+    def test_mismatched_lengths_raise_value_error(self, detector, prompts, answers):
+        """A length mismatch raises a ValueError before the backend gets a request."""
+        with pytest.raises(ValueError, match="Number of prompts must match number of answers"):
+            detector.predict_prompt_batch(prompts, answers)
+        detector.rag.detect_hallucinations_batch.assert_not_called()
+
+    def test_matched_lengths_return_spans_in_input_order(self, detector):
+        """Equal lengths give one spans result for each pair, in input order."""
+        detector.rag.detect_hallucinations_batch.return_value = [HALLUCINATED_RESULT, CLEAN_RESULT]
+        answers = ["Paris is the capital of Germany.", "Berlin is the capital of Germany."]
+
+        results = detector.predict_prompt_batch(["c1", "c2"], answers, output_format="spans")
+
+        assert len(results) == 2
+        assert [span["text"] for span in results[0]] == ["Germany"]
+        assert results[1] == []
+
+    def test_matched_lengths_return_tokens_in_input_order(self, detector):
+        """Equal lengths give one tokens result for each pair, in input order."""
+        detector.rag.detect_hallucinations_batch.return_value = [HALLUCINATED_RESULT, CLEAN_RESULT]
+        answers = ["Paris is the capital of Germany.", "Berlin is the capital of Germany."]
+
+        results = detector.predict_prompt_batch(["c1", "c2"], answers, output_format="tokens")
+
+        assert [len(tokens) for tokens in results] == [6, 6]
+        assert any(token["pred"] == 1 for token in results[0])
+        assert all(token["pred"] == 0 for token in results[1])
+
+    def test_short_backend_result_raises_instead_of_truncating(self, detector):
+        """The call raises a ValueError when the backend returns fewer results than answers."""
+        detector.rag.detect_hallucinations_batch.return_value = [CLEAN_RESULT]
+
+        with pytest.raises(ValueError, match="shorter"):
+            detector.predict_prompt_batch(["c1", "c2"], ["a1", "a2"], output_format="spans")