diff --git a/CHANGELOG.md b/CHANGELOG.md
index 4b4a0ae..5be0cf1 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -24,7 +24,8 @@ All notable changes to Tome are documented here. Format loosely follows
   so books of that type are stored under it, e.g. `Manga/Berserk/…` or
   `Comics/Saga/…`. Empty by default, so nothing moves on upgrade. Existing
   books move only when you run Reorganize in Library Health; a library scan
-  takes a matching top-level folder as a hint for the book type. (#270)
+  takes a matching top-level folder as a hint for the book type, so a plain
+  CBZ under your own `Manga/` stays manga. (#270)
 - **A layout per book type.** Each book type picks how its books are laid
   out (Admin → Types): Series first (the layout so far, still the default),
   Author, then series (`Harlan Coben/Myron Bolitar/Deal Breaker v01.epub`), or
diff --git a/backend/services/book_types.py b/backend/services/book_types.py
index 54fd481..282d78d 100644
--- a/backend/services/book_types.py
+++ b/backend/services/book_types.py
@@ -118,16 +118,19 @@ def detect_book_type(
     fmt: Optional[str],
     is_manga: bool = False,
     in_chapters_dir: bool = False,
+    by_extension: bool = True,
 ) -> Optional[BookType]:
     """The type an incoming file gets automatically, before it is placed.
 
     Same rule the importers have always used: a chapters/ drop or a
     ComicInfo manga flag → manga, a cbz/cbr → comic, anything else none.
+    ``by_extension=False`` keeps only the explicit manga signal, so the scan
+    hint can decide before the bare cbz/cbr default does.
     """
     slug: Optional[str] = None
     if in_chapters_dir or is_manga:
         slug = "manga"
-    elif (fmt or "").lower() in ("cbz", "cbr"):
+    elif by_extension and (fmt or "").lower() in ("cbz", "cbr"):
         slug = "comic"
     if slug is None:
         return None
diff --git a/backend/services/scanner.py b/backend/services/scanner.py
index 8017c50..e660b0d 100644
--- a/backend/services/scanner.py
+++ b/backend/services/scanner.py
@@ -312,9 +312,10 @@ def _create_book_entry(
     """Create the Book + BookFile rows for a file already in place.
 
     ``book_type`` is the type the importer already decided (and placed the
-    file by). Without it the automatic rule applies, and failing that the
-    scan hint: a file under a top-level folder that matches a type's folder
-    or label gets that type (#270).
+    file by). Without it: a manga flag makes it manga; else the scan hint, a
+    top-level folder matching a type's folder or label, gives that type
+    (#270); else the extension rule (cbz/cbr → comic). The hint beats the
+    bare extension so a plain cbz in a user's own ``Manga/`` stays manga.
     """
     book = Book(
         title=meta.get("title", file_path.stem),
@@ -351,12 +352,14 @@ def _create_book_entry(
     ))
     record_ko_hash(db, book.id, ko_partial_md5(file_path), "raw")
 
-    # Book type: as decided by the importer, else the automatic rule, else
-    # the scan hint from the file's top-level folder.
+    # Book type: as decided by the importer, else the manga flag, else the
+    # scan hint from the file's top-level folder, else the extension rule.
+    if book_type is None and folder_hint:
+        book_type = (detect_book_type(db, fmt=fmt, is_manga=bool(meta.get("_is_manga")),
+                                      by_extension=False)
+                     or type_for_top_folder(db, folder_hint))
     if book_type is None:
         book_type = detect_book_type(db, fmt=fmt, is_manga=bool(meta.get("_is_manga")))
-    if book_type is None and folder_hint:
-        book_type = type_for_top_folder(db, folder_hint)
     if book_type is not None:
         book.book_type_id = book_type.id
 
diff --git a/tests/test_type_folders.py b/tests/test_type_folders.py
index 8952d52..f0458f9 100644
--- a/tests/test_type_folders.py
+++ b/tests/test_type_folders.py
@@ -10,7 +10,8 @@
   Library Health lists them, reorganize dry run lists the move, the real run
   performs it and updates the stored path
 - scan hint: a file under a top-level folder matching a type's folder or
-  label gets that type, unless the cbz/manga rule already decided
+  label gets that type; it beats the bare cbz → comic default but not a
+  manga flag
 - two types sharing one folder
 """
 from pathlib import Path
@@ -329,13 +330,31 @@ def test_folder_matching_type_assigns_type(self, db, dirs, types, monkeypatch):
         # The scan never moves files
         assert (lib / "manga/Frieren/Frieren v01.epub").is_file()
 
-    def test_rule_wins_over_hint(self, db, dirs, types, monkeypatch):
+    def test_hint_beats_cbz_default(self, db, dirs, types, monkeypatch):
         lib = dirs["lib"]
         _fake_meta(monkeypatch, "backend.services.scanner.extract_metadata", {})
         (lib / "Manga/Saga").mkdir(parents=True)
         (lib / "Manga/Saga/Saga v01.cbz").write_bytes(b"cbz")
         scan_library(lib, dirs["covers"], db, workers=1)
-        assert db.query(Book).one().book_type_id == types["comic"].id
+        assert db.query(Book).one().book_type_id == types["manga"].id
+
+    def test_manga_flag_beats_hint(self, db, dirs, types, monkeypatch):
+        lib = dirs["lib"]
+        _fake_meta(monkeypatch, "backend.services.scanner.extract_metadata",
+                   {"Akira v01.cbz": {"title": "Akira v01", "_is_manga": True}})
+        (lib / "Comics/Akira").mkdir(parents=True)
+        (lib / "Comics/Akira/Akira v01.cbz").write_bytes(b"cbz")
+        scan_library(lib, dirs["covers"], db, workers=1)
+        assert db.query(Book).one().book_type_id == types["manga"].id
+
+    def test_cbz_without_matching_folder_stays_comic(self, db, dirs, types, monkeypatch):
+        lib = dirs["lib"]
+        _fake_meta(monkeypatch, "backend.services.scanner.extract_metadata", {})
+        for rel in ("Downloads/Saga/Saga v01.cbz", "Loose.cbz"):
+            (lib / rel).parent.mkdir(parents=True, exist_ok=True)
+            (lib / rel).write_bytes(rel.encode())
+        scan_library(lib, dirs["covers"], db, workers=1)
+        assert {b.book_type_id for b in db.query(Book).all()} == {types["comic"].id}
 
     def test_folder_match_beats_label_match(self, db, dirs, types, monkeypatch):
         # "Comics" is the comic type's label but the book type's folder.
diff --git a/website/src/pages/docs/admin.astro b/website/src/pages/docs/admin.astro
index 9fd0f48..3bc01e3 100644
--- a/website/src/pages/docs/admin.astro
+++ b/website/src/pages/docs/admin.astro
@@ -195,7 +195,8 @@ import { withBase } from '../../lib/path'
   <p>
     <strong>New install:</strong> set the folders before the first scan. A scan never moves
     files, but it uses them as a hint: a file whose top-level folder matches a type's folder or
-    label (case-insensitive) gets that type, unless the CBZ/manga rule already decided it. So a
+    label (case-insensitive) gets that type, even a plain CBZ; only a manga flag in the file
+    itself overrides the folder. So a
     library that already has a <code>Manga/</code> folder keeps those books as manga.
   </p>
   <p>