diff --git a/src/modules/ai/prompts/schemas/rag-chat.schema.ts b/src/modules/ai/prompts/schemas/rag-chat.schema.ts
index 3756b6e..1009cbf 100644
--- a/src/modules/ai/prompts/schemas/rag-chat.schema.ts
+++ b/src/modules/ai/prompts/schemas/rag-chat.schema.ts
@@ -1,6 +1,9 @@
+import { z } from 'zod';
+
+export const RagChatOutputSchema = z.object({
+  answer: z.string().min(1).max(5000),
+});
+
 export const RAG_CHAT_OUTPUT_SCHEMA_DESC = `{
-  "answer": "你的回答内容",
-  "citations": [
-    { "title": "引用的知识点标题", "snippet": "引用的原文片段" }
-  ]
+  "answer": "你的回答内容"
 }`;
diff --git a/src/modules/rag-chat/rag-chat.service.ts b/src/modules/rag-chat/rag-chat.service.ts
index 7adf096..9b5e0fc 100644
--- a/src/modules/rag-chat/rag-chat.service.ts
+++ b/src/modules/rag-chat/rag-chat.service.ts
@@ -2,6 +2,7 @@ import { Injectable, NotFoundException, Logger, Optional } from '@nestjs/common'
 import { PrismaService } from '../../infrastructure/database/prisma.service';
 import { ContentSafetyService } from '../content-safety/content-safety.service';
 import { AiGatewayService } from '../ai/gateway/ai-gateway.service';
+import { RagChatOutputSchema } from '../ai/prompts/schemas/rag-chat.schema';
 
 const MAX_CONTEXT_CHARS = 4000;
 
@@ -75,6 +76,7 @@ export class RagChatService {
           promptVersion: 'v1',
           messages,
           maxTokens: 2048,
+          outputSchema: RagChatOutputSchema,
         });
         this.logger.log(`AI Gateway response: parsed=${!!resp.parsed}, keys=${resp.parsed ? Object.keys(resp.parsed).join(',') : 'null'}, raw=${JSON.stringify(resp.parsed).substring(0, 300)}`);
         reply = resp.parsed?.answer ?? String(resp.parsed?.content ?? '抱歉，AI 暂时无法生成回答。');