diff --git a/plugins/douyu/main.py b/plugins/douyu/main.py
index 389d71d..d71c7b3 100644
--- a/plugins/douyu/main.py
+++ b/plugins/douyu/main.py
@@ -1284,7 +1284,7 @@ class DouyuPlugin(MessagePluginInterface):
         user_prompt = (
             "请输出一段适合放在日报图片上半部分的弹幕总结，要求：\n"
             "1. 先用 1 段总述直播氛围与主线。\n"
-            "2. 再用 3-5 条要点总结观众关注点、情绪变化、反复出现的梗。\n"
+            "2. 再用 5 条要点总结观众关注点、情绪变化、反复出现的梗、节奏变化和额外反馈，每条只写一句。\n"
             "3. 语言像运营复盘，简洁自然。\n"
             "4. 不要写标题，不要写“根据数据”。\n\n"
             f"主播：{meta.get('nickname') or meta.get('room_name') or meta.get('room_id')}\n"
diff --git a/plugins/douyu/report_template.py b/plugins/douyu/report_template.py
index da06d65..8ffbada 100644
--- a/plugins/douyu/report_template.py
+++ b/plugins/douyu/report_template.py
@@ -39,12 +39,90 @@ def _split_summary_blocks(danmu_summary: str) -> tuple[str, List[str]]:
     return lead, insight_items
 
 
+def _normalize_summary_bullets(payload: Dict[str, Any], items: List[str], target_count: int = 5) -> List[str]:
+    normalized = [str(item or "").strip() for item in items if str(item or "").strip()]
+    if len(normalized) >= target_count:
+        return normalized[:target_count]
+
+    top_terms = [str(item.get("term") or "").strip() for item in (payload.get("top_terms", []) or []) if str(item.get("term") or "").strip()]
+    merged_templates = [str(item.get("text") or "").strip() for item in (payload.get("merged_templates", []) or []) if str(item.get("text") or "").strip()]
+    peak_buckets = payload.get("peak_buckets", []) or []
+    representative_messages = payload.get("representative_messages", []) or []
+
+    supplements: List[str] = []
+    if top_terms:
+        supplements.append(f"讨论焦点比较集中，弹幕反复围绕 {'、'.join(top_terms[:5])} 展开。")
+    if merged_templates:
+        sample_templates = "；".join(text[:24] for text in merged_templates[:3])
+        supplements.append(f"复读和共识梗比较强，重复内容主要集中在 {sample_templates}。")
+    if peak_buckets:
+        top_bucket = peak_buckets[0]
+        bucket_terms = [str(term.get('term') or '').strip() for term in (top_bucket.get("top_terms", []) or []) if str(term.get('term') or '').strip()]
+        if bucket_terms:
+            supplements.append(
+                f"高峰时段出现在 {str(top_bucket.get('start_time') or '')[-8:-3]} 前后，话题明显偏向 {'、'.join(bucket_terms[:4])}。"
+            )
+    if representative_messages:
+        supplements.append("代表性发言里既有操作反馈，也有玩梗调侃和情绪宣泄，互动意愿比较强。")
+
+    existing = set(normalized)
+    for item in supplements:
+        if item not in existing:
+            normalized.append(item)
+            existing.add(item)
+        if len(normalized) >= target_count:
+            break
+    return normalized[:target_count]
+
+
+def _build_template_items(payload: Dict[str, Any], limit: int = 8) -> List[str]:
+    items: List[str] = []
+    seen = set()
+
+    def push(text: str, suffix: str = "") -> None:
+        value = str(text or "").strip()
+        if not value:
+            return
+        normalized_key = value
+        if normalized_key in seen:
+            return
+        seen.add(normalized_key)
+        items.append(f"{value}{suffix}".strip())
+
+    for item in (payload.get("merged_templates", []) or [])[:6]:
+        text = str(item.get("text") or "").strip()
+        count = int(item.get("count", 0) or 0)
+        if text:
+            push(text[:72], f"（{count}次）")
+
+    for item in (payload.get("repeated_messages", []) or [])[:6]:
+        text = str(item.get("text") or "").strip()
+        count = int(item.get("count", 0) or 0)
+        if text:
+            push(text[:72], f"（{count}次）")
+
+    for item in (payload.get("burst_terms", []) or [])[:6]:
+        text = str(item.get("text") or "").strip()
+        count = int(item.get("count", 0) or 0)
+        if text:
+            push(text[:36], f"（爆发 {count} 次）")
+
+    for item in (payload.get("top_terms", []) or [])[:6]:
+        term = str(item.get("term") or "").strip()
+        count = int(item.get("count", 0) or 0)
+        if term:
+            push(term, f"（{count}次提及）")
+
+    return items[:limit]
+
+
 def _render_insight_cards(items: List[str]) -> str:
     labels = ["主线", "情绪", "梗点", "节奏", "反馈", "补充"]
     blocks = []
     for idx, item in enumerate(items[:6]):
+        extra_class = " full-span" if len(items[:6]) % 2 == 1 and idx == len(items[:6]) - 1 else ""
         blocks.append(
-            '<div class="insight-card">'
+            f'<div class="insight-card{extra_class}">'
             f'<div class="insight-kicker">{_escape(labels[idx] if idx < len(labels) else "观察")}</div>'
             f'<div class="insight-text">{_escape(item)}</div>'
             "</div>"
@@ -103,12 +181,7 @@ def render_daily_report_html(
         _render_metric_card("30+等级用户", operator.get("high_room_level_user_count", 0), "高等级老观众"),
     ])
 
-    merged_templates = payload.get("merged_templates", []) or []
-    template_items = [
-        f"{str(item.get('text') or '').strip()[:72]}（{int(item.get('count', 0) or 0)}次）"
-        for item in merged_templates[:5]
-        if str(item.get("text") or "").strip()
-    ]
+    template_items = _build_template_items(payload)
     top_active_users = payload.get("operator_metrics", {}).get("top_active_users", []) or []
     active_user_items = []
     for item in top_active_users[:10]:
@@ -121,6 +194,7 @@ def render_daily_report_html(
             active_user_items.append(f"{nickname} | {message_count}条")
 
     lead_summary, danmu_bullets = _split_summary_blocks(danmu_summary)
+    danmu_bullets = _normalize_summary_bullets(payload, danmu_bullets, target_count=5)
 
     html_doc = f"""<html>
 <head>
@@ -314,6 +388,9 @@ def render_daily_report_html(
         border: 1px solid rgba(125, 145, 186, 0.14);
         min-height: 110px;
     }}
+    .insight-card.full-span {{
+        grid-column: 1 / -1;
+    }}
     .insight-kicker {{
         color: #2b59ff;
         font-size: 12px;