diff --git a/.gitignore b/.gitignore
index b481110..8a8d457 100644
--- a/.gitignore
+++ b/.gitignore
@@ -90,6 +90,7 @@ data/excel_exports/
 # 本機 QA / 頁面快取
 data/*_cache/
 data/ai_automation_smoke_history.jsonl
+data/ai_automation/sitewide_visual_qa_latest.json
 
 # 上傳檔案
 web/static/uploads/
diff --git a/docs/AI_INTELLIGENCE_MODULE_SOT.md b/docs/AI_INTELLIGENCE_MODULE_SOT.md
index c2047e6..ac44d71 100644
--- a/docs/AI_INTELLIGENCE_MODULE_SOT.md
+++ b/docs/AI_INTELLIGENCE_MODULE_SOT.md
@@ -111,6 +111,7 @@
 - 2026-07-03 起高可見 AI surface HTML readback 必須納入 runtime AI smoke：`/api/ai-automation/surface-html-readback` 會 read-only 檢查 10 個 AI observability template contract，`/api/ai-automation/smoke` 需包含 `AI surface HTML readback` 檢查，`/api/ai-automation/scheduled-health-summary` 需輸出 `ai_surface_html_readback` family；此檢查不得寫 DB、不得送通知、不得要求 browser，失敗時下一個機器動作是 `repair_ai_surface_html_contract`。
 - 2026-07-03 起整站 UI/UX 專業化必須由 Sitewide UI/UX Agent 管理：`/api/ai-automation/sitewide-ui-ux-agent` 會掃描真正頁面模板、排除 partial/components/base layouts，輸出高優先頁、compact guardrail、raw engineering issue 與 next machine action；`/api/ai-automation/sitewide-ui-ux-repair-package` 會產生 no-write controlled repair package。當前 Agent 判定為 55 個頁面模板、31 個高優先模板、29 個已具 compact guardrail、0 個待專業化、0 個 raw engineering leak，Sitewide Agent 狀態為 `ok`，repair package 為 `no_op`。四批 controlled UI/UX repair 已完成 AI observability、AI 建議、PPT/CI/CD/業績/比價/成長/供貨風險高優先 surface；下一個機器動作為 `keep_sitewide_ui_ux_agent_monitoring`，後續只接受 Agent 新增 issue 或更高階 visual/mobile QA 的證據驅動修復。
 - 2026-07-03 起 Sitewide UI/UX Agent 歸零狀態必須納入 AI automation runtime monitoring：`/api/ai-automation/smoke` 需包含 `Sitewide UI/UX Agent readback` 檢查，要求 `issue_surface_count=0`、`repair_package_status=no_op`、`repair_action_count=0`；`/api/ai-automation/scheduled-health-summary` 需輸出 `sitewide_ui_ux_agent` family；`/metrics` 需以 `momo_ai_automation_scheduled_health_family_status{family="sitewide_ui_ux_agent",...}` 暴露 scheduled family。此監控不得寫 DB、不得送通知、不得要求人工審核。
+- 2026-07-03 起整站 desktop/tablet/mobile visual QA 必須進入 AI automation runtime monitoring：`scripts/check_responsive_overflow.js --artifact-output data/ai_automation/sitewide_visual_qa_latest.json` 會輸出機器可讀 artifact；`/api/ai-automation/sitewide-visual-qa-readback` 只讀該 artifact 並回報 route_count、viewport_count、failed_count、overflow_issue_count、stale 與 next_machine_action；`/api/ai-automation/smoke` 需包含 `Sitewide visual QA readback`，`/api/ai-automation/scheduled-health-summary` 需輸出 `sitewide_visual_qa` family，`/metrics` 需暴露 `momo_ai_automation_scheduled_health_family_status{family="sitewide_visual_qa",...}`。readback 不啟動 browser、不寫 DB、不送通知；artifact 生成才需要 Playwright/瀏覽器。
 - 2026-07-02 起 `/ai_intelligence` 商品明細與單品作戰詳情的四格價格證據必須可測：PChome 價格、MOMO 參考價、差距、可信度需以 `data-evidence` 固定，並以 `aria-label="價格證據"` 對應可掃描區塊；候選待確認或缺資料只能顯示「候選待確認 / 待補」，不得捏造價格或讓使用者打開 raw payload 才知道判斷依據。
 - 2026-07-02 起 `/ai_intelligence` 必須是密集 AI 工作台，不得退回大段文字說明頁：首屏與明細可見內容只保留短狀態、數字、四格證據與下一步按鈕；KPI note、benchmark detail、alert 副句、策略說明、decision copy、來源長句與單品 reason list 不得佔用第一層視覺。`tests/test_ai_intelligence_text_density_guardrails.py` 必須鎖住 `data-density-guardrail="compact-ai-workbench"`、短任務文案、detail meta 與 hidden explanatory copy。
 - 2026-07-02 起 `/observability/overview` 也必須採密集 AI 觀測工作台：首屏以 `data-density-guardrail="compact-observability-workbench"`、`AI 觀測 / 風險優先 / 下一步` 與 golden signals 先呈現狀態、數字與操作入口；hero lede、signal note、route desc、host meta 與資料來源長句不得佔用第一層視覺。`tests/test_observability_text_density_guardrails.py` 必須鎖住 compact marker 與 hidden explanatory copy。
diff --git a/docs/guides/external_professional_benchmark.md b/docs/guides/external_professional_benchmark.md
index 271d606..7ef8a91 100644
--- a/docs/guides/external_professional_benchmark.md
+++ b/docs/guides/external_professional_benchmark.md
@@ -84,6 +84,7 @@ Baymard 的商品頁與比較 UX 研究強調：使用者需要清楚的 product
 - Route HTML readback rollout: high-visibility AI observability surfaces 已新增 Flask route-level rendered HTML guard；`tests/test_admin_observability_routes.py::test_high_visibility_ai_surfaces_route_html_readback_keeps_compact_contract` 鎖住 10 個 route 的 compact density marker 與 non-raw engineering wording。
 - Runtime HTML readback rollout: `/api/ai-automation/surface-html-readback` 與 AI smoke 已接入 high-visibility AI surface contract readback；`tests/test_ai_automation_smoke_service.py` 鎖住 10 個 surface、退化偵測、smoke check 與 scheduled health family。
 - Sitewide UI/UX Agent rollout: `/api/ai-automation/sitewide-ui-ux-agent` 掃描真正頁面模板並產生整站專業化 inventory；`/api/ai-automation/sitewide-ui-ux-repair-package` 會輸出 no-write controlled repair package，把主流產品頁原則轉成可排序、可驗證的修復項。四批 controlled repair 已把 AI 建議、PPT 預覽、服務更新監控、當日業績、商品/EDM 舊入口、成長報表、PChome 商品監控、比價決策台、業績作戰分析與供貨風險頁群納入 compact workbench guardrails，Agent 基線收斂到 29 個 compact guardrail / 0 個待專業化 surface。
+- Sitewide visual QA rollout: `scripts/check_responsive_overflow.js --artifact-output data/ai_automation/sitewide_visual_qa_latest.json` 會把 desktop / tablet / mobile overflow QA 輸出為 artifact；`/api/ai-automation/sitewide-visual-qa-readback`、`Sitewide visual QA readback` 與 `sitewide_visual_qa` scheduled family 會把最新視覺 QA 狀態納入 AI automation monitoring。
 
 ## 下一步 TODO 候選
 
@@ -103,6 +104,7 @@ Baymard 的商品頁與比較 UX 研究強調：使用者需要清楚的 product
    - 已完成: `/observability/rag_queries` 首屏以「知識命中 / 省模閉環 / 回饋學習」短標籤與四個核心數字呈現 AI 知識召回狀態。
    - 已完成: `/observability/promotion_review` 首屏以「AI 例外 / 去重守門 / 防污染」短標籤與四個核心數字呈現 AI 晉升例外狀態。
    - 已完成: `/observability/ppt_audit_history` 首屏以「預覽就緒 / 視覺審核 / 修復閉環」短標籤呈現 AI 視覺 QA 產線狀態。
+   - 已完成: Sitewide visual QA artifact readback 會把 desktop / tablet / mobile overflow 結果接入 AI smoke、scheduled health summary 與 `/metrics`。
 
 ## 參考來源
 
diff --git a/docs/guides/pchome_ai_automation_priority_backlog.md b/docs/guides/pchome_ai_automation_priority_backlog.md
index 98d577f..68853d5 100644
--- a/docs/guides/pchome_ai_automation_priority_backlog.md
+++ b/docs/guides/pchome_ai_automation_priority_backlog.md
@@ -179,9 +179,14 @@
   - 此檢查不寫 DB、不發通知、不要求 browser；失敗時 next machine action 是 `repair_ai_surface_html_contract`。
 - 整站 UI/UX Agent inventory 已完成:
   - `/api/ai-automation/sitewide-ui-ux-agent` 會掃描真正頁面模板，排除 partial/components/base layouts。
-  - 當前基線: 55 個頁面模板、31 個高優先模板、14 個 compact guardrail、15 個待專業化、0 個 raw engineering leak。
+  - 當前基線: 55 個頁面模板、31 個高優先模板、29 個 compact guardrail、0 個待專業化 surface、0 個 raw engineering leak。
   - `/api/ai-automation/sitewide-ui-ux-repair-package` 會產生 no-write controlled repair package，優先修復高曝光頁，且不寫 DB、不執行 shell、不讀 secrets。
-  - 第一批 controlled UI/UX repair 已套用到 `ai_history` 與 `ai_recommend`，整站 Agent 已回讀 issue count 從 17 降為 15。
+  - 四批 controlled UI/UX repair 已完成 AI observability、AI 建議、PPT/CI/CD/業績/比價/成長/供貨風險高優先 surface，整站 repair package 目前為 `no_op`。
+- 整站 visual/mobile QA artifact readback 已完成:
+  - `scripts/check_responsive_overflow.js --artifact-output data/ai_automation/sitewide_visual_qa_latest.json` 會把 desktop / tablet / mobile overflow QA 輸出為 machine-readable artifact。
+  - `/api/ai-automation/sitewide-visual-qa-readback` 只讀 artifact，回報 route_count、viewport_count、failed_count、overflow_issue_count、stale 與 next_machine_action。
+  - `/api/ai-automation/smoke` 已加入 `Sitewide visual QA readback`，scheduled health summary 已加入 `sitewide_visual_qa` family，`/metrics` 會輸出對應 Prometheus family status。
+  - readback 不寫 DB、不送通知、不啟動 browser；artifact 生成才需要 Playwright/瀏覽器，避免把 visual QA 變成產品 API 的重型副作用。
 
 已完成 / 下一步，必須照順序:
 
@@ -196,6 +201,8 @@
    - 已完成: 高可見 AI surface readback 已接入 runtime smoke 與 scheduled health summary。
 4. 啟動整站 UI/UX Agent，將所有頁面依主流專業 guardrails 排序並產生 controlled repair package。
    - 已完成: Sitewide UI/UX Agent inventory 與 repair package API。
+5. 將 desktop / tablet / mobile visual QA 變成 AI automation artifact 與 runtime readback。
+   - 已完成: Sitewide visual QA artifact schema、readback API、AI smoke、scheduled family 與 metrics family。
 
 完成標準:
 
diff --git a/docs/memory/current_execution_queue_20260524.md b/docs/memory/current_execution_queue_20260524.md
index ad0fb4c..9c7d0d7 100644
--- a/docs/memory/current_execution_queue_20260524.md
+++ b/docs/memory/current_execution_queue_20260524.md
@@ -174,6 +174,7 @@
 - 表格與圖表只允許在局部容器橫向滾動，不可造成整頁無限延伸。
 - 2026-07-03 起以 Sitewide UI/UX Agent 排序全站專業化工作，不再靠人工逐頁猜測；目前 55 個頁面模板 / 31 個高優先頁中，compact guardrail 已由 14 提升到 29，待專業化 surface 由 15 降到 0。第四批已完成 vendor stockout 剩餘五頁，repair package 進入 `no_op`，下一步轉為 production monitoring 與 desktop/mobile visual QA。
 - 2026-07-03 起 Sitewide UI/UX Agent 已接入 AI automation smoke、scheduled health summary 與 `/metrics` family；之後若整站 UI/UX 退化，主線入口是 `Sitewide UI/UX Agent readback` / `sitewide_ui_ux_agent` family，而不是人工逐頁盤點。
+- 2026-07-03 起 desktop/tablet/mobile visual QA artifact 也接入 AI automation：`scripts/check_responsive_overflow.js --artifact-output data/ai_automation/sitewide_visual_qa_latest.json` 產生 Playwright overflow artifact，`/api/ai-automation/sitewide-visual-qa-readback`、`Sitewide visual QA readback` 與 `sitewide_visual_qa` scheduled family 會讀取最新結果。後續若手機/桌面 overflow 或 artifact stale，主線入口是重新跑 visual QA artifact 並修復 failed routes，不再靠人工截圖逐頁判斷。
 
 ## 8. 效能與可觀測性
 
diff --git a/routes/README.md b/routes/README.md
index d3e4494..b74c760 100644
--- a/routes/README.md
+++ b/routes/README.md
@@ -14,7 +14,7 @@
 |------|------|----------|
 | `dashboard_routes.py` | 商品看板首頁 | `/` |
 | `sales_routes.py` | 業績分析與 ABC 明細 | `/sales_analysis`, `/growth_analysis`, `/abc_analysis/detail`, `/api/sales_analysis/*` |
-| `system_public_routes.py` | 無 prefix 公開系統頁與監控 | `/health`, `/metrics`, `/ai_automation_smoke`, `/api/ai-automation/smoke*`, `/api/ai-automation/scheduled-health-summary`, `/api/ai-automation/surface-html-readback`, `/api/ai-automation/surface-html-repair-package`, `/api/ai-automation/sitewide-ui-ux-agent`, `/api/ai-automation/sitewide-ui-ux-repair-package`, `/settings`, `/system_settings`, `/logs`, `/api/logs`, `/api/backup` |
+| `system_public_routes.py` | 無 prefix 公開系統頁與監控 | `/health`, `/metrics`, `/ai_automation_smoke`, `/api/ai-automation/smoke*`, `/api/ai-automation/scheduled-health-summary`, `/api/ai-automation/surface-html-readback`, `/api/ai-automation/surface-html-repair-package`, `/api/ai-automation/sitewide-ui-ux-agent`, `/api/ai-automation/sitewide-ui-ux-repair-package`, `/api/ai-automation/sitewide-visual-qa-readback`, `/settings`, `/system_settings`, `/logs`, `/api/logs`, `/api/backup` |
 | `system_routes.py` | 內部系統維護 API | `/api/system/*` |
 | `edm_routes.py` | EDM 與節慶儀表板 | `/edm`, `/festival` |
 | `monthly_routes.py` | 月結分析 | `/monthly_summary_analysis`, `/api/monthly_summary_data` |
diff --git a/routes/system_public_routes.py b/routes/system_public_routes.py
index b07a79f..eda04d4 100644
--- a/routes/system_public_routes.py
+++ b/routes/system_public_routes.py
@@ -624,6 +624,16 @@ def ai_automation_sitewide_ui_ux_repair_package_api():
     return jsonify(build_sitewide_ui_ux_repair_package(limit=max(1, min(limit or 12, 50))))
 
 
+@system_public_bp.route('/api/ai-automation/sitewide-visual-qa-readback')
+@login_required
+def ai_automation_sitewide_visual_qa_readback_api():
+    """Read-only sitewide desktop/tablet/mobile visual QA artifact readback."""
+    from services.ai_surface_html_readback_service import build_sitewide_visual_qa_readback
+
+    max_age_hours = request.args.get('max_age_hours', 26, type=int)
+    return jsonify(build_sitewide_visual_qa_readback(max_age_hours=max(1, min(max_age_hours or 26, 168))))
+
+
 @system_public_bp.route('/api/ai-automation/smoke/history/export')
 @login_required
 def ai_automation_smoke_history_export():
diff --git a/scripts/check_responsive_overflow.js b/scripts/check_responsive_overflow.js
index d84e3da..1408a86 100755
--- a/scripts/check_responsive_overflow.js
+++ b/scripts/check_responsive_overflow.js
@@ -110,6 +110,7 @@ function parseArgs(argv) {
     maxOverflow: 1,
     screenshotDir: '',
     screenshotAll: false,
+    artifactOutput: '',
     json: false,
   };
 
@@ -139,6 +140,8 @@ function parseArgs(argv) {
       options.screenshotDir = argv[++i];
     } else if (arg === '--screenshot-all') {
       options.screenshotAll = true;
+    } else if (arg === '--artifact-output') {
+      options.artifactOutput = argv[++i];
     } else if (arg === '--json') {
       options.json = true;
     } else if (arg === '--help' || arg === '-h') {
@@ -173,6 +176,7 @@ Options:
   --max-overflow PX           Allowed body overflow, default 1
   --screenshot-dir DIR        Save failure screenshots
   --screenshot-all            Save screenshots for passing routes too
+  --artifact-output PATH      Write machine-readable sitewide visual QA artifact
   --json                      Print JSON summary
 `);
 }
@@ -315,6 +319,39 @@ function printResult(result) {
   }
 }
 
+function buildArtifact(options, results) {
+  const failed = results.filter((result) => !result.passed);
+  const routes = Array.from(new Set(results.map((result) => result.route))).sort();
+  const viewports = Array.from(new Set(results.map((result) => result.viewport))).sort();
+  return {
+    policy: 'sitewide_responsive_overflow_guard_v1',
+    generated_at: new Date().toISOString(),
+    base_url: options.baseUrl,
+    max_overflow_px: options.maxOverflow,
+    route_count: routes.length,
+    viewport_count: viewports.length,
+    summary: {
+      result_count: results.length,
+      pass_count: results.length - failed.length,
+      failed_count: failed.length,
+      overflow_issue_count: failed.filter((result) => String(result.error || '').toLowerCase().includes('overflow')).length,
+      visual_offender_count: failed.filter((result) => String(result.error || '').toLowerCase().includes('offender')).length,
+    },
+    routes,
+    viewports,
+    results,
+  };
+}
+
+function writeArtifact(options, artifact) {
+  if (!options.artifactOutput) {
+    return;
+  }
+  const outputPath = path.resolve(options.artifactOutput);
+  fs.mkdirSync(path.dirname(outputPath), { recursive: true });
+  fs.writeFileSync(outputPath, `${JSON.stringify(artifact, null, 2)}\n`, 'utf-8');
+}
+
 async function main() {
   const options = parseArgs(process.argv.slice(2));
   const { chromium } = requirePlaywright();
@@ -395,6 +432,9 @@ async function main() {
     await closeWithTimeout(browser);
   }
 
+  const artifact = buildArtifact(options, results);
+  writeArtifact(options, artifact);
+
   if (options.json) {
     console.log(JSON.stringify(results, null, 2));
   }
diff --git a/services/ai_automation_smoke_service.py b/services/ai_automation_smoke_service.py
index 4cf314f..29dc8c9 100644
--- a/services/ai_automation_smoke_service.py
+++ b/services/ai_automation_smoke_service.py
@@ -212,6 +212,8 @@ def build_scheduled_automation_health_summary(
     surface_details = surface_readback.get("details") or {}
     sitewide_readback = _find_check(source_result, "Sitewide UI/UX Agent readback")
     sitewide_details = sitewide_readback.get("details") or {}
+    visual_qa_readback = _find_check(source_result, "Sitewide visual QA readback")
+    visual_qa_details = visual_qa_readback.get("details") or {}
     smoke_status = source_result.get("status") or ("warning" if latest_history else "warning")
     freshness_family = _history_freshness_family(
         latest_history,
@@ -300,6 +302,30 @@ def build_scheduled_automation_health_summary(
                 "writes_database": False,
             },
         },
+        {
+            "key": "sitewide_visual_qa",
+            "label": "Sitewide visual QA",
+            "status": visual_qa_readback.get("status") or "warning",
+            "summary": visual_qa_readback.get("summary") or "Sitewide visual QA has no latest check.",
+            "next_machine_action": (
+                "keep_sitewide_visual_qa_monitoring"
+                if visual_qa_readback.get("status") == "ok"
+                else "run_sitewide_visual_qa_and_publish_artifact"
+            ),
+            "details": {
+                "policy": visual_qa_details.get("policy"),
+                "result_count": int(visual_qa_details.get("result_count") or 0),
+                "route_count": int(visual_qa_details.get("route_count") or 0),
+                "viewport_count": int(visual_qa_details.get("viewport_count") or 0),
+                "failed_count": int(visual_qa_details.get("failed_count") or 0),
+                "overflow_issue_count": int(visual_qa_details.get("overflow_issue_count") or 0),
+                "visual_offender_count": int(visual_qa_details.get("visual_offender_count") or 0),
+                "artifact_path": visual_qa_details.get("artifact_path"),
+                "artifact_generated_at": visual_qa_details.get("artifact_generated_at"),
+                "stale": bool(visual_qa_details.get("stale")),
+                "writes_database": False,
+            },
+        },
         freshness_family,
         {
             "key": "daily_summary_delivery",
@@ -938,6 +964,61 @@ def _sitewide_ui_ux_agent_check() -> Dict[str, Any]:
         )
 
 
+def _sitewide_visual_qa_check() -> Dict[str, Any]:
+    """Read-only sentinel for latest desktop/tablet/mobile visual QA artifact."""
+    try:
+        from services.ai_surface_html_readback_service import build_sitewide_visual_qa_readback
+
+        readback = build_sitewide_visual_qa_readback()
+        summary = readback.get("summary") or {}
+        status = readback.get("status") or "warning"
+        failed_count = int(summary.get("failed_count") or 0)
+        route_count = int(summary.get("route_count") or 0)
+        viewport_count = int(summary.get("viewport_count") or 0)
+        summary_text = (
+            f"Sitewide visual QA ok，{route_count} routes x {viewport_count} viewports 無 overflow 退化"
+            if status == "ok"
+            else (
+                f"Sitewide visual QA 需更新或修復：failed={failed_count}、"
+                f"routes={route_count}、stale={bool(summary.get('stale'))}"
+            )
+        )
+        return _check(
+            "Sitewide visual QA readback",
+            status,
+            summary_text,
+            {
+                "policy": readback.get("policy"),
+                "result_count": int(summary.get("result_count") or 0),
+                "route_count": route_count,
+                "viewport_count": viewport_count,
+                "pass_count": int(summary.get("pass_count") or 0),
+                "failed_count": failed_count,
+                "overflow_issue_count": int(summary.get("overflow_issue_count") or 0),
+                "visual_offender_count": int(summary.get("visual_offender_count") or 0),
+                "artifact_path": readback.get("artifact_path"),
+                "artifact_exists": bool(readback.get("artifact_exists")),
+                "artifact_generated_at": readback.get("artifact_generated_at"),
+                "age_hours": summary.get("age_hours"),
+                "stale": bool(summary.get("stale")),
+                "next_machine_action": readback.get("next_machine_action"),
+                "writes_database": False,
+                "writes_database_count": 0,
+                "primary_human_gate_count": 0,
+            },
+        )
+    except Exception as exc:
+        return _check(
+            "Sitewide visual QA readback",
+            "critical",
+            f"Sitewide visual QA readback 無法執行：{exc}",
+            {
+                "writes_database": False,
+                "primary_human_gate_count": 0,
+            },
+        )
+
+
 def collect_ai_automation_smoke(*, record_history: bool = True, history_limit: int = 20) -> Dict[str, Any]:
     checks: List[Dict[str, Any]] = [
         _event_router_check(),
@@ -949,6 +1030,7 @@ def collect_ai_automation_smoke(*, record_history: bool = True, history_limit: i
         _pchome_controlled_apply_drift_monitor_check(),
         _ai_surface_html_readback_check(),
         _sitewide_ui_ux_agent_check(),
+        _sitewide_visual_qa_check(),
     ]
     worst = max(checks, key=lambda item: STATUS_RANK.get(item["status"], 2))["status"]
     result = {
diff --git a/services/ai_surface_html_readback_service.py b/services/ai_surface_html_readback_service.py
index e8fbb6f..db1bfaa 100644
--- a/services/ai_surface_html_readback_service.py
+++ b/services/ai_surface_html_readback_service.py
@@ -2,6 +2,8 @@
 
 from __future__ import annotations
 
+import json
+import os
 from dataclasses import dataclass
 from datetime import datetime
 from pathlib import Path
@@ -14,7 +16,13 @@ POLICY = "read_only_ai_surface_html_readback_v1"
 REPAIR_POLICY = "read_only_ai_surface_html_repair_package_v1"
 SITEWIDE_POLICY = "read_only_sitewide_ui_ux_agent_inventory_v1"
 SITEWIDE_REPAIR_POLICY = "read_only_sitewide_ui_ux_repair_package_v1"
+SITEWIDE_VISUAL_QA_POLICY = "read_only_sitewide_visual_qa_readback_v1"
 ROOT = Path(__file__).resolve().parents[1]
+SITEWIDE_VISUAL_QA_ARTIFACT_PATH = os.getenv(
+    "MOMO_SITEWIDE_VISUAL_QA_ARTIFACT",
+    str(ROOT / "data" / "ai_automation" / "sitewide_visual_qa_latest.json"),
+)
+SITEWIDE_VISUAL_QA_MAX_AGE_HOURS = int(os.getenv("MOMO_SITEWIDE_VISUAL_QA_MAX_AGE_HOURS", "26"))
 
 
 @dataclass(frozen=True)
@@ -135,6 +143,143 @@ PROFESSIONAL_GUARDRAIL_MARKERS: tuple[str, ...] = (
 )
 
 
+def _parse_timestamp(value: Any) -> datetime | None:
+    if not value:
+        return None
+    text_value = str(value).strip()
+    if not text_value:
+        return None
+    try:
+        return datetime.fromisoformat(text_value.replace("Z", "+00:00"))
+    except ValueError:
+        return None
+
+
+def _visual_qa_payload_from_results(raw_payload: Any) -> dict[str, Any]:
+    if isinstance(raw_payload, list):
+        return {"results": raw_payload}
+    if isinstance(raw_payload, dict):
+        return raw_payload
+    return {"results": []}
+
+
+def _visual_qa_summary(payload: dict[str, Any]) -> dict[str, Any]:
+    results = list(payload.get("results") or [])
+    failed = [item for item in results if not item.get("passed")]
+    routes = sorted({str(item.get("route") or "") for item in results if item.get("route")})
+    viewports = sorted({str(item.get("viewport") or "") for item in results if item.get("viewport")})
+    overflow_issues = [
+        item for item in failed
+        if "overflow" in str(item.get("error") or "").lower()
+    ]
+    visual_offenders = [
+        item for item in failed
+        if "offender" in str(item.get("error") or "").lower()
+    ]
+    return {
+        "result_count": len(results),
+        "route_count": len(routes),
+        "viewport_count": len(viewports),
+        "pass_count": len(results) - len(failed),
+        "failed_count": len(failed),
+        "overflow_issue_count": len(overflow_issues),
+        "visual_offender_count": len(visual_offenders),
+        "routes": routes,
+        "viewports": viewports,
+    }
+
+
+def build_sitewide_visual_qa_readback(
+    *,
+    artifact_path: Path | str | None = None,
+    artifact_payload: dict[str, Any] | list[dict[str, Any]] | None = None,
+    max_age_hours: int | None = None,
+) -> dict[str, Any]:
+    """Read the latest Playwright responsive QA artifact without running a browser."""
+    path = Path(artifact_path or SITEWIDE_VISUAL_QA_ARTIFACT_PATH)
+    max_age = max(1, int(max_age_hours or SITEWIDE_VISUAL_QA_MAX_AGE_HOURS))
+    read_error = ""
+    payload: dict[str, Any]
+    if artifact_payload is not None:
+        payload = _visual_qa_payload_from_results(artifact_payload)
+        source_kind = "inline_payload"
+        artifact_exists = True
+    else:
+        source_kind = "artifact_file"
+        artifact_exists = path.exists()
+        if not artifact_exists:
+            payload = {"results": []}
+            read_error = f"artifact_not_found:{path}"
+        else:
+            try:
+                payload = _visual_qa_payload_from_results(json.loads(path.read_text(encoding="utf-8")))
+            except (OSError, json.JSONDecodeError) as exc:
+                payload = {"results": []}
+                read_error = str(exc)[:300]
+
+    summary = _visual_qa_summary(payload)
+    generated_at = payload.get("generated_at") or payload.get("completed_at")
+    generated_dt = _parse_timestamp(generated_at)
+    age_hours = None
+    stale = False
+    if generated_dt is not None:
+        now = datetime.now(generated_dt.tzinfo) if generated_dt.tzinfo else datetime.now()
+        age_hours = max((now - generated_dt).total_seconds() / 3600, 0)
+        stale = age_hours > max_age
+
+    status = "ok"
+    if read_error or summary["result_count"] == 0:
+        status = "warning"
+    elif summary["failed_count"] > 0:
+        status = "critical"
+    elif generated_dt is None or stale:
+        status = "warning"
+
+    failed_results = [
+        item for item in list(payload.get("results") or [])
+        if not item.get("passed")
+    ][:12]
+    return {
+        "policy": SITEWIDE_VISUAL_QA_POLICY,
+        "status": status,
+        "version": SYSTEM_VERSION,
+        "generated_at": datetime.now().isoformat(timespec="seconds"),
+        "source_kind": source_kind,
+        "artifact_path": str(path),
+        "artifact_exists": artifact_exists,
+        "artifact_generated_at": generated_at,
+        "summary": {
+            **summary,
+            "age_hours": round(age_hours, 2) if age_hours is not None else None,
+            "max_age_hours": max_age,
+            "stale": stale,
+            "read_error": read_error,
+            "primary_human_gate_count": 0,
+            "writes_database_count": 0,
+        },
+        "failed_results": failed_results,
+        "next_machine_action": (
+            "keep_sitewide_visual_qa_monitoring"
+            if status == "ok"
+            else "run_sitewide_visual_qa_and_publish_artifact"
+        ),
+        "automation_policy": {
+            "primary_flow": "ai_controlled",
+            "manual_review_mode": "exception_only",
+            "machine_verifiable_evidence": True,
+            "primary_human_gate_count": 0,
+        },
+        "safety": {
+            "read_only": True,
+            "writes_database": False,
+            "writes_database_count": 0,
+            "sends_notifications": False,
+            "requires_browser_for_readback": False,
+            "requires_browser_for_artifact_generation": True,
+        },
+    }
+
+
 def _read_surface_html(root: Path, contract: SurfaceContract) -> str:
     return (root / contract.template).read_text(encoding="utf-8")
 
diff --git a/tests/test_ai_automation_metrics.py b/tests/test_ai_automation_metrics.py
index 0e943c7..0971f09 100644
--- a/tests/test_ai_automation_metrics.py
+++ b/tests/test_ai_automation_metrics.py
@@ -107,10 +107,10 @@ def test_system_metrics_exports_scheduled_health_summary():
         Gauge,
         {
             "summary": {
-                "ok": 3,
+                "ok": 4,
                 "warning": 1,
                 "critical": 0,
-                "total": 4,
+                "total": 5,
                 "primary_human_gate_count": 0,
                 "writes_database_count": 0,
             },
@@ -118,12 +118,13 @@ def test_system_metrics_exports_scheduled_health_summary():
                 {"key": "ai_automation_smoke", "status": "ok"},
                 {"key": "pchome_controlled_apply_drift_monitor", "status": "warning"},
                 {"key": "sitewide_ui_ux_agent", "status": "ok"},
+                {"key": "sitewide_visual_qa", "status": "ok"},
             ],
         },
     )
 
     output = generate_latest(registry).decode("utf-8")
-    assert 'momo_ai_automation_scheduled_health_summary_total{status="ok"} 3.0' in output
+    assert 'momo_ai_automation_scheduled_health_summary_total{status="ok"} 4.0' in output
     assert 'momo_ai_automation_scheduled_health_summary_total{status="warning"} 1.0' in output
     assert (
         'momo_ai_automation_scheduled_health_family_status{family="ai_automation_smoke",status="ok"} 1.0'
@@ -137,6 +138,10 @@ def test_system_metrics_exports_scheduled_health_summary():
         'momo_ai_automation_scheduled_health_family_status{family="sitewide_ui_ux_agent",status="ok"} 1.0'
         in output
     )
+    assert (
+        'momo_ai_automation_scheduled_health_family_status{family="sitewide_visual_qa",status="ok"} 1.0'
+        in output
+    )
     assert "momo_ai_automation_scheduled_health_primary_human_gate_count 0.0" in output
     assert "momo_ai_automation_scheduled_health_writes_database_count 0.0" in output
 
diff --git a/tests/test_ai_automation_smoke_service.py b/tests/test_ai_automation_smoke_service.py
index 422fc86..59c7482 100644
--- a/tests/test_ai_automation_smoke_service.py
+++ b/tests/test_ai_automation_smoke_service.py
@@ -38,11 +38,12 @@ def test_collect_ai_automation_smoke_uses_worst_status(monkeypatch):
     monkeypatch.setattr(smoke, "_pchome_controlled_apply_drift_monitor_check", lambda: smoke._check("pchome", "ok", "ok"))
     monkeypatch.setattr(smoke, "_ai_surface_html_readback_check", lambda: smoke._check("surface", "ok", "ok"))
     monkeypatch.setattr(smoke, "_sitewide_ui_ux_agent_check", lambda: smoke._check("sitewide", "ok", "ok"))
+    monkeypatch.setattr(smoke, "_sitewide_visual_qa_check", lambda: smoke._check("visual", "ok", "ok"))
 
     result = smoke.collect_ai_automation_smoke(record_history=False)
 
     assert result["status"] == "critical"
-    assert result["summary"] == {"ok": 7, "warning": 1, "critical": 1, "total": 9}
+    assert result["summary"] == {"ok": 8, "warning": 1, "critical": 1, "total": 10}
 
 
 def test_pchome_controlled_apply_drift_monitor_reports_verified_zero_drift(monkeypatch):
@@ -111,6 +112,7 @@ def test_collect_ai_automation_smoke_persists_recent_history(tmp_path, monkeypat
     monkeypatch.setattr(smoke, "_pchome_controlled_apply_drift_monitor_check", lambda: smoke._check("pchome", "ok", "ok"))
     monkeypatch.setattr(smoke, "_ai_surface_html_readback_check", lambda: smoke._check("surface", "ok", "ok"))
     monkeypatch.setattr(smoke, "_sitewide_ui_ux_agent_check", lambda: smoke._check("sitewide", "ok", "ok"))
+    monkeypatch.setattr(smoke, "_sitewide_visual_qa_check", lambda: smoke._check("visual", "ok", "ok"))
 
     first = smoke.collect_ai_automation_smoke(history_limit=5)
     second = smoke.collect_ai_automation_smoke(history_limit=5)
@@ -166,7 +168,7 @@ def test_scheduled_automation_health_summary_reads_history_without_side_effects(
         json.dumps({
             "generated_at": datetime.now().isoformat(timespec="seconds"),
             "status": "ok",
-            "summary": {"ok": 9, "warning": 0, "critical": 0, "total": 9},
+            "summary": {"ok": 10, "warning": 0, "critical": 0, "total": 10},
             "checks": [
                 {
                     "name": "PChome 受控落地 drift monitor",
@@ -206,6 +208,23 @@ def test_scheduled_automation_health_summary_reads_history_without_side_effects(
                         "repair_package_status": "no_op",
                         "repair_action_count": 0,
                     },
+                },
+                {
+                    "name": "Sitewide visual QA readback",
+                    "status": "ok",
+                    "summary": "Sitewide visual QA ok，12 routes x 3 viewports 無 overflow 退化",
+                    "details": {
+                        "policy": "read_only_sitewide_visual_qa_readback_v1",
+                        "result_count": 36,
+                        "route_count": 12,
+                        "viewport_count": 3,
+                        "failed_count": 0,
+                        "overflow_issue_count": 0,
+                        "visual_offender_count": 0,
+                        "artifact_path": "data/ai_automation/sitewide_visual_qa_latest.json",
+                        "artifact_generated_at": datetime.now().isoformat(timespec="seconds"),
+                        "stale": False,
+                    },
                 }
             ],
         }, ensure_ascii=False) + "\n",
@@ -224,7 +243,7 @@ def test_scheduled_automation_health_summary_reads_history_without_side_effects(
     )
     assert summary["policy"] == "read_only_ai_automation_scheduled_health_summary"
     assert summary["status"] == "ok"
-    assert summary["summary"]["total"] == 6
+    assert summary["summary"]["total"] == 7
     assert summary["summary"]["primary_human_gate_count"] == 0
     assert summary["summary"]["writes_database_count"] == 0
     assert pchome_family["status"] == "ok"
@@ -242,6 +261,13 @@ def test_scheduled_automation_health_summary_reads_history_without_side_effects(
     assert sitewide_family["status"] == "ok"
     assert sitewide_family["details"]["issue_surface_count"] == 0
     assert sitewide_family["details"]["repair_package_status"] == "no_op"
+    visual_family = next(
+        item for item in summary["families"]
+        if item["key"] == "sitewide_visual_qa"
+    )
+    assert visual_family["status"] == "ok"
+    assert visual_family["details"]["route_count"] == 12
+    assert visual_family["details"]["failed_count"] == 0
     assert summary["scheduled_outputs"]["telegram_send_in_preview"] is False
     assert summary["scheduled_outputs"]["writes_database"] is False
     assert summary["automation_policy"]["primary_flow"] == "ai_controlled"
@@ -256,7 +282,7 @@ def test_scheduled_automation_health_summary_can_use_current_smoke_without_recor
     current_smoke = {
         "generated_at": "2026-07-02T12:00:00",
         "status": "critical",
-        "summary": {"ok": 8, "warning": 0, "critical": 1, "total": 9},
+        "summary": {"ok": 9, "warning": 0, "critical": 1, "total": 10},
         "checks": [
             {
                 "name": "PChome 受控落地 drift monitor",
@@ -281,6 +307,19 @@ def test_scheduled_automation_health_summary_can_use_current_smoke_without_recor
                     "repair_package_status": "no_op",
                     "repair_action_count": 0,
                 },
+            },
+            {
+                "name": "Sitewide visual QA readback",
+                "status": "ok",
+                "summary": "Sitewide visual QA ok，12 routes x 3 viewports 無 overflow 退化",
+                "details": {
+                    "route_count": 12,
+                    "viewport_count": 3,
+                    "failed_count": 0,
+                    "overflow_issue_count": 0,
+                    "visual_offender_count": 0,
+                    "stale": False,
+                },
             }
         ],
     }
@@ -307,7 +346,7 @@ def test_scheduled_automation_health_summary_route_returns_compact_payload(tmp_p
         json.dumps({
             "generated_at": datetime.now().isoformat(timespec="seconds"),
             "status": "ok",
-            "summary": {"ok": 8, "warning": 0, "critical": 0, "total": 8},
+            "summary": {"ok": 10, "warning": 0, "critical": 0, "total": 10},
             "checks": [],
         }, ensure_ascii=False) + "\n",
         encoding="utf-8",
@@ -474,6 +513,58 @@ def test_sitewide_ui_ux_agent_reports_non_utf8_templates(tmp_path):
     assert package["safety"]["writes_database"] is False
 
 
+def test_sitewide_visual_qa_readback_consumes_artifact_payload():
+    from datetime import datetime
+    from services.ai_surface_html_readback_service import build_sitewide_visual_qa_readback
+
+    payload = build_sitewide_visual_qa_readback(
+        artifact_payload={
+            "generated_at": datetime.now().isoformat(timespec="seconds"),
+            "results": [
+                {"route": "/ai_intelligence", "viewport": "desktop-1440", "passed": True},
+                {"route": "/ai_intelligence", "viewport": "mobile-390", "passed": True},
+                {"route": "/observability/overview", "viewport": "mobile-390", "passed": True},
+            ],
+        }
+    )
+
+    assert payload["policy"] == "read_only_sitewide_visual_qa_readback_v1"
+    assert payload["status"] == "ok"
+    assert payload["summary"]["result_count"] == 3
+    assert payload["summary"]["route_count"] == 2
+    assert payload["summary"]["viewport_count"] == 2
+    assert payload["summary"]["failed_count"] == 0
+    assert payload["summary"]["primary_human_gate_count"] == 0
+    assert payload["summary"]["writes_database_count"] == 0
+    assert payload["safety"]["requires_browser_for_readback"] is False
+    assert payload["safety"]["requires_browser_for_artifact_generation"] is True
+
+
+def test_sitewide_visual_qa_readback_reports_visual_failures():
+    from datetime import datetime
+    from services.ai_surface_html_readback_service import build_sitewide_visual_qa_readback
+
+    payload = build_sitewide_visual_qa_readback(
+        artifact_payload={
+            "generated_at": datetime.now().isoformat(timespec="seconds"),
+            "results": [
+                {
+                    "route": "/price_comparison",
+                    "viewport": "mobile-390",
+                    "passed": False,
+                    "error": "body horizontal overflow 24px",
+                }
+            ],
+        }
+    )
+
+    assert payload["status"] == "critical"
+    assert payload["summary"]["failed_count"] == 1
+    assert payload["summary"]["overflow_issue_count"] == 1
+    assert payload["next_machine_action"] == "run_sitewide_visual_qa_and_publish_artifact"
+    assert payload["failed_results"][0]["route"] == "/price_comparison"
+
+
 def test_surface_html_readback_check_is_part_of_ai_smoke(monkeypatch):
     from services import ai_automation_smoke_service as smoke
 
@@ -484,6 +575,12 @@ def test_surface_html_readback_check_is_part_of_ai_smoke(monkeypatch):
     monkeypatch.setattr(smoke, "_embedding_queue_check", lambda: smoke._check("embedding", "ok", "ok"))
     monkeypatch.setattr(smoke, "_elephant_hitl_check", lambda: smoke._check("elephant", "ok", "ok"))
     monkeypatch.setattr(smoke, "_pchome_controlled_apply_drift_monitor_check", lambda: smoke._check("pchome", "ok", "ok"))
+    monkeypatch.setattr(smoke, "_sitewide_visual_qa_check", lambda: smoke._check(
+        "Sitewide visual QA readback",
+        "ok",
+        "visual ok",
+        {"route_count": 12, "viewport_count": 3, "failed_count": 0},
+    ))
 
     result = smoke.collect_ai_automation_smoke(record_history=False)
 
@@ -495,12 +592,18 @@ def test_surface_html_readback_check_is_part_of_ai_smoke(monkeypatch):
         item for item in result["checks"]
         if item["name"] == "Sitewide UI/UX Agent readback"
     )
-    assert result["summary"]["total"] == 9
+    visual_check = next(
+        item for item in result["checks"]
+        if item["name"] == "Sitewide visual QA readback"
+    )
+    assert result["summary"]["total"] == 10
     assert surface_check["status"] == "ok"
     assert surface_check["details"]["checked_surface_count"] == 10
     assert sitewide_check["status"] == "ok"
     assert sitewide_check["details"]["issue_surface_count"] == 0
     assert sitewide_check["details"]["repair_package_status"] == "no_op"
+    assert visual_check["status"] == "ok"
+    assert visual_check["details"]["route_count"] == 12
 
 
 def test_surface_html_readback_route_returns_payload(monkeypatch):
@@ -538,6 +641,11 @@ def test_surface_and_sitewide_ui_ux_agent_routes_return_payloads():
             routes.ai_automation_sitewide_ui_ux_repair_package_api.__wrapped__,
             "read_only_sitewide_ui_ux_repair_package_v1",
         ),
+        (
+            "/api/ai-automation/sitewide-visual-qa-readback?max_age_hours=9999",
+            routes.ai_automation_sitewide_visual_qa_readback_api.__wrapped__,
+            "read_only_sitewide_visual_qa_readback_v1",
+        ),
     ]
 
     for path, view_func, policy in route_cases: