diff --git a/docs/reports/Viewclinic_QB_Results.xlsx b/docs/reports/Viewclinic_QB_Results.xlsx
index 12695d3..837e8a0 100644
Binary files a/docs/reports/Viewclinic_QB_Results.xlsx and b/docs/reports/Viewclinic_QB_Results.xlsx differ
diff --git a/scripts/export_qb_excel.py b/scripts/export_qb_excel.py
index 74a181e..bf1ced5 100644
--- a/scripts/export_qb_excel.py
+++ b/scripts/export_qb_excel.py
@@ -1,7 +1,7 @@
 #!/usr/bin/env python3
 """Question Bank 120문항 + 2엔진 실측 결과를 엑셀(xlsx)로 내보낸다.
 
-시트: 요약 / 통합결과 / 병원빈도 / GPT상세 / PPLX상세
+시트: 요약 / 통합결과 / 병원빈도 / TOP20 / GPT상세 / PPLX상세
 """
 import json
 import os
@@ -145,6 +145,54 @@ def main():
     for col, w in zip("ABCD", [16, 30, 13, 13]):
         ws.column_dimensions[col].width = w
 
+    # ── TOP20 (논브랜드 84문항 · 2엔진 합산 랭킹) ─────────
+    ws = wb.create_sheet("TOP20")
+    ws["A1"] = "성형외과 AI Discovery TOP 20 (논브랜드 84문항 · ChatGPT+Perplexity 합산)"
+    ws["A1"].font = Font(bold=True, size=14, color=INK)
+    ws["A2"] = "점수 = 답변 내 등장 순위 가중치(1위 5점 ~ 5위 1점)를 질문·엔진 단위로 합산. 브랜드 직접 질문 36문항은 제외."
+    ws["A2"].font = Font(size=10, color="666B7A")
+    ws.append([])
+    ws.append(["순위", "병원", "AI Discovery 점수", "언급 질문 수(합산)", "GPT 언급", "PPLX 언급", "1순위 등장", "주요 등장 카테고리"])
+    hr = ws.max_row
+    style_header(ws, 8, hr)
+
+    agg = {}
+    for eng_key, data in (("gpt", oa), ("pplx", px)):
+        for q in qs:
+            if brand(q):
+                continue
+            seen = set()
+            for i, c in enumerate([c for c in data[q["id"]].get("clinics", []) if isinstance(c, str)][:5]):
+                key = canon(c)
+                if key in seen:
+                    continue
+                seen.add(key)
+                a = agg.setdefault(key, {"label": "뷰성형외과" if key == "뷰성형외과" else c,
+                                         "score": 0, "gpt": 0, "pplx": 0,
+                                         "first": 0, "cats": {}})
+                a["score"] += 5 - i
+                a[eng_key] += 1
+                if i == 0:
+                    a["first"] += 1
+                cat = CAT_LABEL[q["id"].split("-")[0]]
+                a["cats"][cat] = a["cats"].get(cat, 0) + 1
+
+    ranked = sorted(agg.values(), key=lambda a: (-a["score"], -a["first"], -(a["gpt"] + a["pplx"])))
+    view_rank = next((i + 1 for i, a in enumerate(ranked) if canon(a["label"]) == "뷰성형외과"), None)
+    shown = ranked[:20]
+    if view_rank and view_rank > 20:
+        shown = ranked[:20] + [ranked[view_rank - 1]]
+    for idx, a in enumerate(shown):
+        rank = idx + 1 if idx < 20 else view_rank
+        cats = " · ".join(k for k, _ in sorted(a["cats"].items(), key=lambda x: -x[1])[:3])
+        ws.append([rank, a["label"], a["score"], a["gpt"] + a["pplx"], a["gpt"], a["pplx"], a["first"], cats])
+        if canon(a["label"]) == "뷰성형외과":
+            for c in range(1, 9):
+                ws.cell(row=ws.max_row, column=c).fill = PatternFill("solid", fgColor=WARN)
+                ws.cell(row=ws.max_row, column=c).font = Font(bold=True)
+    for col, w in zip("ABCDEFGH", [6, 30, 16, 16, 10, 10, 10, 40]):
+        ws.column_dimensions[col].width = w
+
     # ── 엔진별 상세 (답변 전문 + 인용) ────────────────────
     for sheet, data in (("GPT상세", oa), ("PPLX상세", px)):
         ws = wb.create_sheet(sheet)