diff --git a/app/models/document.py b/app/models/document.py
index cf6d3cc..dff88c8 100644
--- a/app/models/document.py
+++ b/app/models/document.py
@@ -35,6 +35,7 @@ class Document(Base):
 
     # 2계층: 추출 메타 (OCR 판정/실행)
     extract_meta: Mapped[dict | None] = mapped_column(JSONB, default=dict)
+    ocr_derived: Mapped[bool] = mapped_column(Boolean, nullable=False, default=False)
 
     # 2계층: AI 가공
     ai_summary: Mapped[str | None] = mapped_column(Text)
diff --git a/migrations/277_documents_ocr_derived.sql b/migrations/277_documents_ocr_derived.sql
new file mode 100644
index 0000000..0e0942a
--- /dev/null
+++ b/migrations/277_documents_ocr_derived.sql
@@ -0,0 +1,4 @@
+-- 2026-05-24 PR-Chore-OCR-Column-1 (1/2): documents.ocr_derived boolean 컬럼 추가.
+-- RAG-independent data hygiene. ocr_derived 식별 컬럼 부재 = PR-Eval-V0_2 TBD-O FAILED 원인.
+-- backfill 은 278 별 migration (asyncpg single-statement 제약).
+ALTER TABLE documents ADD COLUMN IF NOT EXISTS ocr_derived boolean DEFAULT false NOT NULL;
diff --git a/migrations/278_documents_ocr_derived_backfill.sql b/migrations/278_documents_ocr_derived_backfill.sql
new file mode 100644
index 0000000..132fbd2
--- /dev/null
+++ b/migrations/278_documents_ocr_derived_backfill.sql
@@ -0,0 +1,5 @@
+-- 2026-05-24 PR-Chore-OCR-Column-1 (2/2): ocr_derived backfill.
+-- Rule R1 단독 (실측 audit 후): extract_meta ? ocr_attempted AND ocr_attempted = true → 8 rows.
+-- R2 (이미지 file_format) 폐기 = 1건 R1 흡수 + 1건 marker 미처리 false.
+-- R3 (marker PDF extract_meta 부재 283 rows) 폐기 = born-digital false positive 위험.
+UPDATE documents SET ocr_derived = true WHERE deleted_at IS NULL AND extract_meta ? 'ocr_attempted' AND (extract_meta->>'ocr_attempted')::boolean = true;