diff --git a/web-ui/src/lib/api/precedent-library.ts b/web-ui/src/lib/api/precedent-library.ts
index 9e0c723..50f4720 100644
--- a/web-ui/src/lib/api/precedent-library.ts
+++ b/web-ui/src/lib/api/precedent-library.ts
@@ -646,11 +646,26 @@ export function useHalachotPending(opts: { limit?: number; search?: string } = {
   });
 }
 
+/** Genuine extraction defects — the rule TEXT is malformed, so it needs re-extraction
+ * (not a chair keep/drop judgment). Mirrors the backend's DEFECT set in
+ * scripts/halacha_panel_approve.py; other flags (e.g. `application` = fact-specific
+ * application) are judgment signals, NOT defects, and the panel deliberately skips them. */
+const EXTRACTION_DEFECT_FLAGS = new Set([
+  "quote_unverified",
+  "truncated_quote",
+  "thin_restatement",
+  "near_duplicate",
+]);
+
 /** A pending item belongs in the "needs extraction fix" segment when it carries a
- * quality flag AND the panel never deliberated it (no round). Everything else —
- * deliberated items and clean items — is a chair-judgment item. (#133 unified queue) */
+ * genuine extraction-DEFECT flag AND the panel never deliberated it (no round).
+ * Everything else — deliberated items, clean items, and items flagged only for
+ * chair judgment (e.g. `application`) — is a chair-judgment item. (#133 unified queue) */
 export function isExtractionFixItem(h: Halacha): boolean {
-  return !h.panel_round && (h.quality_flags?.length ?? 0) > 0;
+  return (
+    !h.panel_round &&
+    (h.quality_flags ?? []).some((f) => EXTRACTION_DEFECT_FLAGS.has(f))
+  );
 }
 
 export function useHalachotByStatus(status: string, limit = 300) {