diff --git a/backend/alembic/versions/0100_drop_plex_item_markers.py b/backend/alembic/versions/0100_drop_plex_item_markers.py
new file mode 100644
index 00000000..03991be6
--- /dev/null
+++ b/backend/alembic/versions/0100_drop_plex_item_markers.py
@@ -0,0 +1,22 @@
+"""plex_items: drop `markers`, a column nothing writes or reads
+
+Intro/credits markers are read live from Plex for the item the player opens (with Plex's `final` flag
+on the block that ends it); this cached copy was never filled — every row is NULL. The downgrade
+restores the empty column.
+"""
+
+import sqlalchemy as sa
+from alembic import op
+
+revision: str = "0100_drop_plex_item_markers"
+down_revision: str | None = "0099_library_states"
+branch_labels: str | None = None
+depends_on: str | None = None
+
+
+def upgrade() -> None:
+    op.drop_column("plex_items", "markers")
+
+
+def downgrade() -> None:
+    op.add_column("plex_items", sa.Column("markers", sa.JSON(), nullable=True))
diff --git a/backend/app/models.py b/backend/app/models.py
index b8b1a4fb..1a4e06db 100644
--- a/backend/app/models.py
+++ b/backend/app/models.py
@@ -1274,9 +1274,6 @@ class PlexItem(Base, TimestampMixin, UpdatedAtMixin):
     # bit-depth/HDR. NULL = never (the backfill's work-list). Only HW-decodable transcode items are
     # probed (the rest are SW-only by codec and need no bit-depth); see `_probe_media_batch`.
     media_probed_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True))
-    # Intro/credit markers: [{"type": "intro"|"credits", "start_s": int, "end_s": int}, ...] — unused:
-    # the player reads them live from Plex (`routes/plex/detail.item_detail`, with Plex's `final`).
-    markers: Mapped[list | None] = mapped_column(JSON)
     added_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True), index=True)
     # Filterable / orderable metadata mirrored from the Plex section listing (no extra API calls;
     # `rating` = Plex audienceRating, which is ~the IMDb score). The arrays back genre/person/studio