diff --git a/packages/report/pdf/templates/report.html.j2 b/packages/report/pdf/templates/report.html.j2
index b9a1b56..c0b2366 100644
--- a/packages/report/pdf/templates/report.html.j2
+++ b/packages/report/pdf/templates/report.html.j2
@@ -3,6 +3,11 @@
 <head>
 <meta charset="utf-8">
 <title>{{ meta.brand }} · {{ meta.title }}</title>
+<!-- CR-05 文件元数据隐式标识（N2'-①）：WeasyPrint 会把 author/description/generator
+     映射进 PDF 文档属性（Author/Subject/Creator）；27 位内容编号随合规出口 publish_asset() 落地，此处不伪造 -->
+<meta name="author" content="美搭（AI 生成内容）">
+<meta name="description" content="本报告由人工智能生成，仅供参考。报告 {{ meta.report_id }} v{{ meta.version }}，生成于 {{ date_text }}">
+<meta name="generator" content="美搭 AI 报告生成服务">
 <style>
 /* ── 页面与页脚：A4 分页，页脚带 AI 标识 + 免责语（合规 §合规与安全方案）── */
 @page {
diff --git a/packages/shared/errors.py b/packages/shared/errors.py
index 15fd251..abf3930 100644
--- a/packages/shared/errors.py
+++ b/packages/shared/errors.py
@@ -80,7 +80,7 @@ ERROR_META: dict[ErrorCode, ErrorMeta] = {
     ErrorCode.BAD_REQUEST: ErrorMeta(400, "请求参数有误", "请检查后重试"),
     ErrorCode.VALIDATION_FAILED: ErrorMeta(422, "请求参数校验未通过", "请按提示修改后重试"),
     ErrorCode.IDEMPOTENCY_CONFLICT: ErrorMeta(
-        409, "相同请求正在处理中", "请稍后查看任务中心，不要重复提交"
+        409, "相同请求正在处理中", "请稍后在首页「报告」中查看"
     ),
     ErrorCode.UPLOAD_REJECTED: ErrorMeta(
         400, "不支持的图片类型或超出大小限制", "请上传 20MB 以内的 JPG/PNG/WebP 图片"
@@ -115,7 +115,7 @@ ERROR_META: dict[ErrorCode, ErrorMeta] = {
     ErrorCode.SMS_DAILY_LIMIT: ErrorMeta(429, "今日发送次数已达上限", "明天再试"),
     ErrorCode.INTERNAL_ERROR: ErrorMeta(500, "服务开小差了", "请稍后重试"),
     ErrorCode.DB_ERROR: ErrorMeta(500, "服务开小差了", "请稍后重试"),
-    ErrorCode.TASK_STATE_INVALID: ErrorMeta(500, "任务状态异常", "请到任务中心重试"),
+    ErrorCode.TASK_STATE_INVALID: ErrorMeta(500, "任务状态异常", "请稍后在首页「报告」中查看"),
     ErrorCode.CONFIG_ERROR: ErrorMeta(500, "服务配置异常", "请联系客服"),
     ErrorCode.NOT_IMPLEMENTED: ErrorMeta(501, "该能力尚未开放"),
     ErrorCode.UPSTREAM_ERROR: ErrorMeta(502, "外部服务暂时不可用", "请稍后重试"),
diff --git a/tests/unit/test_report_export.py b/tests/unit/test_report_export.py
index 75c49b6..31c4e1d 100644
--- a/tests/unit/test_report_export.py
+++ b/tests/unit/test_report_export.py
@@ -9,6 +9,8 @@
 
 from __future__ import annotations
 
+import re
+import zlib
 from collections.abc import AsyncIterator
 from contextlib import asynccontextmanager
 from datetime import UTC, datetime
@@ -24,6 +26,7 @@ from apps.api.main import create_app
 from apps.api.routers.reports import report_export_service_dep
 from packages.report.dimensions import DIMENSIONS
 from packages.report.export_service import ExportFile, ExportSubmission
+from packages.report.pdf import renderer as pdf_renderer
 from packages.report.pdf.renderer import (
     ReportPdfMeta,
     _reject_all_urls,
@@ -115,6 +118,61 @@ def test_url_fetcher_rejects_everything_including_file_scheme() -> None:
             _reject_all_urls(url)
 
 
+# --- PDF 元数据 AI 声明（CR-05 隐式标识，R4 N2'-①） -----------------------
+
+
+def _render_template_html() -> str:
+    """用渲染器同款 Jinja 环境+上下文形状渲染 HTML（与 render_report_pdf 内部一致）。"""
+    template = pdf_renderer._env.get_template("report.html.j2")
+    return template.render(
+        meta=META,
+        date_text=META.generated_at.strftime("%Y年%m月%d日"),
+        summary="",
+        sections=[],
+        dimensions=[],
+        core_dimensions=[],
+        ok_count=0,
+    )
+
+
+def _decompressed_streams(pdf: bytes) -> list[bytes]:
+    """解出 PDF 里全部 FlateDecode 流——WeasyPrint 把 Info 字典放进压缩对象流，
+    裸字节 grep 不到 /Author 等标记（zlib 为标准库，不新增依赖）。"""
+    out: list[bytes] = []
+    for m in re.finditer(rb"stream\r?\n", pdf):
+        start = m.end()
+        end = pdf.find(b"endstream", start)
+        if end < 0:
+            continue
+        try:
+            out.append(zlib.decompress(pdf[start:end]))
+        except zlib.error:
+            continue
+    return out
+
+
+def test_report_html_declares_ai_generation_meta() -> None:
+    html = _render_template_html()
+    assert '<meta name="author" content="美搭（AI 生成内容）">' in html
+    assert 'name="description"' in html
+    # description 携带报告号/版本/日期（ReportPdfMeta 现成数据，不新增字段）
+    assert "报告 101 v1" in html
+    assert "生成于 2026年08月29日" in html
+
+
+def test_rendered_pdf_carries_ai_metadata() -> None:
+    # WeasyPrint 把 meta author/description/generator 映射进 PDF 信息字典
+    # （Author/Subject/Creator）；值编码为带 BOM 的 UTF-16BE 十六进制文本串。
+    pdf = render_report_pdf(_sample_result(), meta=META)
+    joined = b"\n".join(_decompressed_streams(pdf))
+    for marker in (b"/Author", b"/Subject", b"/Creator"):
+        assert marker in joined, f"PDF 信息字典缺 {marker!r}"
+    author_hex = "feff" + "美搭（AI 生成内容）".encode("utf-16-be").hex()
+    assert author_hex.encode("ascii") in joined
+    desc_hex = "feff" + "本报告由人工智能生成，仅供参考。报告 101 v1".encode("utf-16-be").hex()
+    assert desc_hex.encode("ascii") in joined
+
+
 # --- 路由接线（假服务，不连库） --------------------------------------------