diff --git a/test_unstructured/file_utils/test_filetype.py b/test_unstructured/file_utils/test_filetype.py
index e942714373..1f90b33d13 100644
--- a/test_unstructured/file_utils/test_filetype.py
+++ b/test_unstructured/file_utils/test_filetype.py
@@ -1105,15 +1105,16 @@ def and_it_uses_character_detection_to_correct_a_wrong_encoding_arg_for_file_pat
         assert len(text_head) == 4096
         assert text_head.startswith("Iwan Roberts\nRoberts celebrating after")
 
-    def but_not_to_correct_a_wrong_encoding_arg_for_a_file_like_object_open_in_binary_mode(self):
-        """Fails silently in this case, returning empty string."""
+    def and_it_uses_character_detection_to_correct_a_wrong_encoding_arg_for_a_file_like_object_open_in_binary_mode(self):
         with open(example_doc_path("norwich-city.txt"), "rb") as f:
             file = io.BytesIO(f.read())
         ctx = _FileTypeDetectionContext(file=file, encoding="utf_32_be")
 
         text_head = ctx.text_head
 
-        assert text_head == ""
+        assert isinstance(text_head, str)
+        assert len(text_head) == 4063
+        assert text_head.startswith("Iwan Roberts\nRoberts celebrating after")
 
     def and_it_grabs_the_first_4k_chars_from_binary_file_for_textual_type_differentiation(self):
         with open(example_doc_path("norwich-city.txt"), "rb") as f:
@@ -1147,9 +1148,6 @@ def it_accommodates_a_utf_32_encoded_file_path(self):
         assert len(text_head) == 188
         assert text_head.startswith("This is a test document to use for unit tests.\n\n    Doyle")
 
-    # TODO: this fails because `.text_head` ignores decoding errors on a file open for binary
-    # reading. Probably better if it used chardet in that case as it does for a file-path.
-    @pytest.mark.xfail(reason="WIP", raises=AssertionError, strict=True)
     def and_it_accommodates_a_utf_32_encoded_file_like_object(self):
         with open(example_doc_path("fake-text-utf-32.txt"), "rb") as f:
             file = io.BytesIO(f.read())
diff --git a/unstructured/file_utils/filetype.py b/unstructured/file_utils/filetype.py
index a12a6249e9..3219259056 100644
--- a/unstructured/file_utils/filetype.py
+++ b/unstructured/file_utils/filetype.py
@@ -673,19 +673,21 @@ def text_head(self) -> str:
 
         Raises:
             UnicodeDecodeError if file cannot be read as text.
+            UnprocessableEntityError if file encoding cannot be determined.
         """
-        # TODO: only attempts fallback character-set detection for file-path case, not for
-        # file-like object case. Seems like we should do both.
-
         if file := self._file_arg:
             file.seek(0)
             content = file.read(4096)
             file.seek(0)
-            return (
-                content
-                if isinstance(content, str)
-                else content.decode(encoding=self.encoding, errors="ignore")
-            )
+            
+            if isinstance(content, str):
+                return content
+                
+            try:
+                return content.decode(encoding=self.encoding)
+            except UnicodeDecodeError:
+                encoding, _ = detect_file_encoding(file=content)
+                return content.decode(encoding=encoding, errors="ignore")
 
         file_path = self.file_path
         assert file_path is not None  # -- guaranteed by `._validate` --