diff --git a/scripts/rank_tracker.py b/scripts/rank_tracker.py
index cdd8c19..34fd3e3 100644
--- a/scripts/rank_tracker.py
+++ b/scripts/rank_tracker.py
@@ -42,7 +42,7 @@ def get_published_articles() -> List[Dict]:
                     "title": article.title or topic.title,
                     "status": article.status,
                     "topic_title": topic.title,
-                    "field": topic.field or "",
+                    "field": topic.field_name or "",
                 })
             return articles
         finally:
@@ -67,7 +67,7 @@ def _build_search_queries(article: Dict) -> List[str]:
             if 4 <= len(p) <= 25:
                 queries.append(p)
 
-    if field:
+    if field and len(field) >= 4:
         queries.append(field[:20])
 
     return list(set(q for q in queries if len(q) >= 4))[:5]
@@ -101,8 +101,47 @@ def _generate_keywords_for_article(article: Dict) -> List[str]:
     return []
 
 
+def _call_llm(prompt: str, max_tokens: int = 1000) -> str:
+    try:
+        from app.core.nvidia_client import call_llm
+        return call_llm(prompt, temperature=0.3, max_tokens=max_tokens) or ""
+    except Exception as e:
+        logger.warning(f"LLM 调用失败: {e}")
+        return ""
+
+
+def _check_ai_citation(article: Dict, keyword: str) -> Dict:
+    title = article.get("title") or article.get("topic_title", "")
+    prompt = f"""你是一个 AI 搜索结果评估器。请判断以下文章是否可能被 AI 搜索引擎（如 Bing AI、百度 AI 搜索、Perplexity）引用。
+
+文章标题：{title}
+搜索关键词：{keyword}
+
+请用以下格式返回：
+被引用：是/否
+引擎：Bing AI, 百度 AI 搜索, Perplexity（逗号分隔，若被引用）
+证据：简短的屏幕抓取或推理依据
+
+只返回以上三行，不要多余内容。"""
+    response = _call_llm(prompt, max_tokens=500)
+    cited = False
+    ai_sources = []
+    if response:
+        cited = "是" in response.split("\n")[0] if "\n" in response else "是" in response
+        if cited:
+            for line in response.split("\n"):
+                if line.startswith("引擎"):
+                    sources = [s.strip() for s in line.replace("引擎：", "").replace("引擎:", "").split(",") if s.strip()]
+                    ai_sources = sources
+    return {
+        "ai_cited": cited,
+        "ai_source": ",".join(ai_sources) if ai_sources else None,
+        "citation_snippet": response[:300] if cited else None,
+    }
+
+
 def check_rankings(article: Dict, keywords: List[str], engine: str = "bing") -> List[Dict]:
-    """检查文章关键词在搜索引擎的排名"""
+    """检查文章关键词在搜索引擎的排名，并检测 AI 搜索引用"""
     title = article.get("title") or article.get("topic_title", "")
     article_id = article.get("id", "")
     topic_id = article.get("topic_id", "")
@@ -122,6 +161,10 @@ def check_rankings(article: Dict, keywords: List[str], engine: str = "bing") ->
                     url_found = url[:200]
                     break
 
+            ai_info = _check_ai_citation(article, keyword)
+            if ai_info.get("ai_cited"):
+                logger.info(f"  [AI] '{keyword}' → 被 AI 搜索引用 ({ai_info.get('ai_source', '未知')})")
+
             results.append({
                 "article_id": article_id,
                 "topic_id": topic_id,
@@ -130,8 +173,8 @@ def check_rankings(article: Dict, keywords: List[str], engine: str = "bing") ->
                 "search_engine": engine,
                 "position": position,
                 "url_found": url_found,
-                "ai_cited": False,
-                "ai_source": None,
+                "ai_cited": ai_info.get("ai_cited", False),
+                "ai_source": ai_info.get("ai_source"),
             })
 
             logger.info(f"  [{engine}] '{keyword}' → {'#' + str(position) if position else '未上榜'}"