diff --git a/pyiceberg/expressions/visitors.py b/pyiceberg/expressions/visitors.py
index 209a57a325..51b47db991 100644
--- a/pyiceberg/expressions/visitors.py
+++ b/pyiceberg/expressions/visitors.py
@@ -1527,12 +1527,14 @@ def eval(self, file: DataFile) -> bool:
         Returns: false if the file may contain any row that doesn't match
                     the expression, true otherwise.
         """
-        if file.record_count <= 0:
-            # Older version don't correctly implement record count from avro file and thus
-            # set record count -1 when importing avro tables to iceberg tables. This should
-            # be updated once we implemented and set correct record count.
+        if file.record_count == 0:
             return ROWS_MUST_MATCH
 
+        if file.record_count < 0:
+            # Older versions set the record count to -1 when importing Avro tables.
+            # Treat an unknown count conservatively rather than as an empty file.
+            return ROWS_MIGHT_NOT_MATCH
+
         return visit(self.expr, _StrictMetricsEvaluationVisitor(self.struct, file))
 
 
diff --git a/tests/expressions/test_evaluator.py b/tests/expressions/test_evaluator.py
index f4c2f89491..57c06af71e 100644
--- a/tests/expressions/test_evaluator.py
+++ b/tests/expressions/test_evaluator.py
@@ -1523,7 +1523,7 @@ def test_metrics_evaluator_record_count_short_circuits() -> None:
     )
     assert _InclusiveMetricsEvaluator(schema, EqualTo("x", 10)).eval(negative_record_file) is ROWS_MIGHT_MATCH
     assert _StrictMetricsEvaluator(schema, EqualTo("x", 10)).eval(zero_record_file) is ROWS_MUST_MATCH
-    assert _StrictMetricsEvaluator(schema, EqualTo("x", 10)).eval(negative_record_file) is ROWS_MUST_MATCH
+    assert _StrictMetricsEvaluator(schema, EqualTo("x", 10)).eval(negative_record_file) is ROWS_MIGHT_NOT_MATCH
 
 
 def test_strict_not(schema_data_file: Schema, strict_data_file_1: DataFile) -> None: