diff --git a/backend/app/data/csv_grid.py b/backend/app/data/csv_grid.py
index 83e6177c..7fb8ece7 100644
--- a/backend/app/data/csv_grid.py
+++ b/backend/app/data/csv_grid.py
@@ -2,9 +2,12 @@
 和网格（Grid / 逐行流），之后的执行器、核对、说明、溯源、按期累积全部复用，不另写一套 CSV 的解析（否则会和配方
 语言分叉，违背「一种说法」）。
 
-**怎么认出是 CSV。** 看内容不看扩展名（配方的几个入口只拿得到原件字节）：不是 zip（xlsx）、不是旧版 .xls 的 OLE
-文件头、前 64 KB 里没有 NUL 字节（图片、PDF 这类二进制文件都有），就按 CSV 读。编码识别、分隔符嗅探复用 tabular
-读 CSV 的那一段（UTF-8 / GB18030 / 带 BOM 的 UTF-16，逗号、分号、制表符、竖线）。
+**怎么认出是 CSV。** 和简单导入（tabular._CSV）一样按扩展名定：只有 .csv / .tsv 按 CSV 读，并且内容得是文字
+（不是 zip、不是旧版 .xls 的 OLE 文件头、前 64 KB 里没有 NUL 字节）。以前只看内容，企业系统常见的「HTML 表格冒充
+.xls」、误传的文本文件都会被拆成一张叫「CSV」的工作表去起草、核对，不再像以前那样明确拒收；现在别的扩展名照旧交给
+xlsx_scan 拒收，内容是网页或 XML 的另给一句能照着做的话。编码识别复用 tabular 读 CSV 的那一段（UTF-8 / GB18030 /
+带 BOM 的 UTF-16）；.tsv 一律按制表符分隔（同 tabular），.csv 嗅探逗号、分号、制表符、竖线。扫描时定下的分隔符
+记进工作表的 path（CSV_PATH 后面跟分隔符），之后读格子用同一个，不再各嗅各的。
 
 **坐标沿用「工作表!A1」。** 细案原写「格子坐标记成 R{行}C{列}」，这里没这么做（见交付说明的偏离项）：坐标在执行器、
 核对、修复按钮、框选、溯源、裁判摘录、前端网格里一共有十来处解析和拼写，引入第二种写法就要处处认两种，漏一处就是
@@ -28,13 +31,23 @@ from typing import Any, Iterator
 from app.data.recipe_types import GridCell
 from app.data.xlsx_scan import Bounds, SheetScan, UnsupportedTable, WorkbookScan
 
-__all__ = ["CSV_PATH", "CSV_SHEET", "MAX_COLS", "MAX_ROWS", "CsvTable", "cell_value", "looks_like_csv", "parse",
-           "rows", "scan", "scan_any"]
+__all__ = ["CSV_PATH", "CSV_SHEET", "CSV_SUFFIXES", "MAX_COLS", "MAX_ROWS", "CsvTable", "cell_value", "is_csv_sheet",
+           "looks_like_csv", "looks_like_markup", "parse", "rows", "scan", "scan_any", "sheet_delimiter"]
 
 #: CSV 的「工作表名」：配方按它认（SheetMatch.name），每月的文件名不同也认得出
 CSV_SHEET = "CSV"
-#: SheetScan.path 的取值：读取层据此知道这张「工作表」要从 CSV 里读，不去压缩包里找 XML
+#: SheetScan.path 的开头：读取层据此知道这张「工作表」要从 CSV 里读，不去压缩包里找 XML。后面跟着扫描时定下的分隔符
+#: （"csv:,"、"csv:\t"）；只有「csv:」的（这一版之前的暂存区）读的时候再嗅一次
 CSV_PATH = "csv:"
+#: 按 CSV 读的扩展名，和简单导入（tabular._CSV）一致
+CSV_SUFFIXES = (".csv", ".tsv")
+#: 嗅探认的分隔符
+_DELIMITERS = ",;\t|"
+#: 内容是网页或 XML（「HTML 表格冒充 .xls」、Excel 2003 的 XML 表格）时的说法
+MARKUP_MESSAGE = ("无法打开该文件：它其实是网页或 XML 格式（有的系统导出的「Excel」文件是这样），不是 Excel 工作簿。"
+                  "请用 Excel 打开后另存为 .xlsx 或 .csv 再上传")
+NOT_TEXT_MESSAGE = ("无法按 CSV 读取这个文件：扩展名是 CSV，内容却不是文字（可能是图片、PDF 之类的文件改了扩展名）。"
+                    "请确认上传的文件，或在 Excel 中另存为 .csv 后重新上传")
 #: 和 Excel 的上限一致：超过的 CSV 用 Excel 打开也装不下，交给模型的格子坐标也写不出来
 MAX_ROWS = 1_048_576
 MAX_COLS = 16_384
@@ -57,6 +70,24 @@ def looks_like_csv(raw: bytes) -> bool:
     return b"\x00" not in raw[:_SNIFF]
 
 
+def looks_like_markup(raw: bytes) -> bool:
+    """内容是网页或 XML：去掉 BOM 和开头的空白后以「<」开头。真的工作簿以 zip（PK）或 OLE 文件头开头，不会是这样。"""
+    head = raw[:4096]
+    if head[:2] in (b"\xff\xfe", b"\xfe\xff"):
+        try:
+            text = head.decode("utf-16", errors="ignore")
+        except UnicodeDecodeError:
+            return False
+    else:
+        text = head.decode("latin-1")
+    return text.lstrip("\ufeff\xef\xbb\xbf \t\r\n").startswith("<")
+
+
+def _suffix(filename: str | None) -> str:
+    name = str(filename or "").rsplit("/", 1)[-1]
+    return "." + name.rsplit(".", 1)[-1].lower() if "." in name else ""
+
+
 def cell_value(text: str) -> Any:
     """CSV 的一个字段 → 格子的值：空 / 只有空白是 None（空格），写法毫无歧义的整数、小数转成数，其余原样（文字）。"""
     t = text.strip()
@@ -79,15 +110,17 @@ class CsvTable:
     delimiter: str
 
 
-def parse(raw: bytes) -> CsvTable:
-    """原件 → 记录。编码、分隔符照 tabular 读 CSV 的做法；字段太长、引号不配对读不了时抛 UnsupportedTable。"""
+def parse(raw: bytes, delimiter: str | None = None) -> CsvTable:
+    """原件 → 记录。编码照 tabular 读 CSV 的做法；delimiter 不给时嗅探（同 tabular）。字段太长、引号不配对读不了时抛
+    UnsupportedTable。"""
     from app.data import tabular
 
     text = tabular._decode_csv(raw)
-    try:
-        delimiter = csv.Sniffer().sniff(text[:8192], delimiters=",;\t|").delimiter
-    except csv.Error:
-        delimiter = ","       # 嗅不出来就按最常见的来
+    if delimiter is None:
+        try:
+            delimiter = csv.Sniffer().sniff(text[:8192], delimiters=_DELIMITERS).delimiter
+        except csv.Error:
+            delimiter = ","       # 嗅不出来就按最常见的来
     records: list[list[str]] = []
     try:
         for record in csv.reader(io.StringIO(text, newline=""), delimiter=delimiter):
@@ -109,11 +142,22 @@ def _cells(table: CsvTable) -> Iterator[tuple[int, list[tuple[int, GridCell]]]]:
             yield r, out
 
 
-def scan(raw: bytes) -> WorkbookScan:
-    """CSV 原件 → 扫描结果：一张可见工作表（名字 CSV_SHEET、路径 CSV_PATH），边界、非空格数、逐行非空格数的游程。
-    没有合并区、隐藏行列、公式、错误值、离群格（CSV 里每一格都是写出来的，没有残留格式）。"""
-    table = parse(raw)
-    sheet = SheetScan(name=CSV_SHEET, state="visible", path=CSV_PATH)
+def is_csv_sheet(sheet: Any) -> bool:
+    """这张「工作表」是 CSV 适配出来的（path 以 CSV_PATH 开头）。"""
+    return str(getattr(sheet, "path", "") or "").startswith(CSV_PATH)
+
+
+def sheet_delimiter(sheet: Any) -> str | None:
+    """扫描时定下的分隔符（记在 path 里）；这一版之前的暂存区没记，返回 None（读的时候再嗅）。"""
+    tail = str(getattr(sheet, "path", "") or "")[len(CSV_PATH):]
+    return tail if len(tail) == 1 and tail in _DELIMITERS else None
+
+
+def scan(raw: bytes, *, delimiter: str | None = None) -> WorkbookScan:
+    """CSV 原件 → 扫描结果：一张可见工作表（名字 CSV_SHEET、路径 CSV_PATH 加分隔符），边界、非空格数、逐行非空格数的
+    游程。没有合并区、隐藏行列、公式、错误值、离群格（CSV 里每一格都是写出来的，没有残留格式）。"""
+    table = parse(raw, delimiter)
+    sheet = SheetScan(name=CSV_SHEET, state="visible", path=CSV_PATH + table.delimiter)
     lo_r = lo_c = hi_r = hi_c = 0
     runs: list[tuple[int, int, int]] = []
     for r, cells in _cells(table):
@@ -134,12 +178,20 @@ def scan(raw: bytes) -> WorkbookScan:
     return WorkbookScan(sheets=[sheet], format="csv")
 
 
-def scan_any(raw: bytes) -> WorkbookScan:
-    """配方导入的扫描入口：CSV 走这里的适配，其余照旧交给 xlsx_scan（读不了照旧拒收，文案不变）。"""
+def scan_any(raw: bytes, filename: str | None) -> WorkbookScan:
+    """配方导入的扫描入口：.csv / .tsv 走这里的适配（.tsv 按制表符分隔），其余照旧交给 xlsx_scan（读不了照旧拒收，
+    文案不变）；内容是网页或 XML 的另给一句能照着做的话（MARKUP_MESSAGE）。扩展名是 CSV、内容却是工作簿（xlsx 改了
+    扩展名）的照工作簿读。"""
     from app.data import xlsx_scan
 
-    if looks_like_csv(raw):
-        return scan(raw)
+    suffix = _suffix(filename)
+    if suffix in CSV_SUFFIXES:
+        if looks_like_csv(raw):
+            return scan(raw, delimiter="\t" if suffix == ".tsv" else None)
+        if raw[:4] != b"PK\x03\x04":
+            raise UnsupportedTable(NOT_TEXT_MESSAGE)
+    elif looks_like_markup(raw):
+        raise UnsupportedTable(MARKUP_MESSAGE)
     return xlsx_scan.scan(raw)
 
 
diff --git a/backend/app/data/recipe_imports.py b/backend/app/data/recipe_imports.py
index e715a08c..2901abba 100644
--- a/backend/app/data/recipe_imports.py
+++ b/backend/app/data/recipe_imports.py
@@ -177,7 +177,8 @@ class Pipeline:
     recipe_sha256: Callable[[Recipe], str]
     canonical: Callable[[Recipe], dict[str, Any]]
     apply_patch: Callable[[dict[str, Any], list[dict[str, Any]]], dict[str, Any]]
-    scan: Callable[[bytes], WorkbookScan]
+    #: (原件, 文件名) → 扫描结果。文件名定按 CSV 读还是按工作簿读（csv_grid.scan_any）
+    scan: Callable[[bytes, str], WorkbookScan]
     scan_to_json: Callable[[WorkbookScan], dict[str, Any]]
     scan_from_json: Callable[[dict[str, Any]], WorkbookScan]
     read_grid: Callable[..., Grid]
@@ -474,7 +475,7 @@ def _open_book(p: Pipeline, raw: bytes, scan: WorkbookScan) -> _Book:
 
 def _book_of(p: Pipeline, snap: _Snap) -> _Book:
     raw = _read_raw(snap.raw_sha256)
-    scan = p.scan_from_json(snap.scan) if snap.scan else p.scan(raw)
+    scan = p.scan_from_json(snap.scan) if snap.scan else p.scan(raw, snap.file_name)
     return _open_book(p, raw, scan)
 
 
@@ -772,7 +773,7 @@ class _Staged:
 
 def _stage_work(p: Pipeline, raw: bytes, filename: str) -> _Staged:
     """首次导入（或切换）的暂存：扫描、网格预览、规则起草（带干跑）、对草稿做一次静态校验。"""
-    scan = p.scan(raw)
+    scan = p.scan(raw, filename)
     book = _open_book(p, raw, scan)
     dry = _Capture(_dry_fn(p, book, filename, {}))
     d = p.draft(scan, book.grids, filename, validate=lambda data, f, o: p.validate(data, facts=f, origin=o),
@@ -860,7 +861,7 @@ async def stage_upload(session: AsyncSession, *, raw: bytes, filename: str, name
 def _replay_work(p: Pipeline, raw: bytes, current: dict[str, Any], filename: str, snap: _Snap,
                  *, evaluate: bool) -> tuple[_Staged, _Eval | None]:
     """上传新一期 / 修改配方：扫描（或取存下的扫描）、按现行配方的写法取系统发现，修改配方时再评估一遍。"""
-    scan = p.scan_from_json(snap.scan) if snap.scan else p.scan(raw)
+    scan = p.scan_from_json(snap.scan) if snap.scan else p.scan(raw, filename)
     book = _open_book(p, raw, scan)
     facts = p.detect_facts(book.grids, scan, recipe=current)
     staged = _Staged(scan_json=p.scan_to_json(scan), previews=[p.preview(g) for g in book.grids.values()],
@@ -1860,7 +1861,7 @@ def _trial_work(p: Pipeline, snap: _Snap, work: dict[str, Any], facts: DraftFact
     if parsed is None or problems:
         return _TrialRun(parsed=parsed, problems=list(problems))
     raw = _read_raw(snap.raw_sha256)
-    scan = p.scan_from_json(snap.scan) if snap.scan else p.scan(raw)
+    scan = p.scan_from_json(snap.scan) if snap.scan else p.scan(raw, snap.file_name)
     path.parent.mkdir(parents=True, exist_ok=True)
     # 先写 .tmp- 再改名：半成品归启动清理（_sweep_tmp）管，发布、回收不会把半个库当成试运行库
     tmp = path.parent / f"{path.name}{raw_store.TMP_MARK}{secrets.token_hex(6)}"
diff --git a/backend/app/data/xlsx_cells.py b/backend/app/data/xlsx_cells.py
index c270f8bb..69f93fad 100644
--- a/backend/app/data/xlsx_cells.py
+++ b/backend/app/data/xlsx_cells.py
@@ -72,10 +72,11 @@ class Workbooks:
         self._formulas: Any = None
         self._csv: Any = None
 
-    def csv(self) -> Any:
-        """CSV 原件解析好的记录（csv_grid.CsvTable），第一次要读时才解析，同一次执行里复用。"""
+    def csv(self, delimiter: str | None = None) -> Any:
+        """CSV 原件解析好的记录（csv_grid.CsvTable），第一次要读时才解析，同一次执行里复用。delimiter 是扫描时定下的
+        分隔符（csv_grid.sheet_delimiter），不给时再嗅一次。"""
         if self._csv is None:
-            self._csv = csv_grid.parse(self.raw)
+            self._csv = csv_grid.parse(self.raw, delimiter)
         return self._csv
 
     def values(self) -> Any:
@@ -133,9 +134,10 @@ def sheet_rows(books: Workbooks, sheet: SheetScan, *, min_row: int, max_row: int
     行数和每行的宽度都一样。文件内容写坏了（见 _READ_ERRORS）抛 UnsupportedTable，日志只记异常类型。
     """
     try:
-        if sheet.path == csv_grid.CSV_PATH:
-            # CSV 适配出来的「工作表」：从解析好的记录里取，口径同下（只给非空格、整行为空的不给）
-            yield from csv_grid.rows(books.csv(), min_row=min_row, max_row=max_row, min_col=min_col, max_col=max_col)
+        if csv_grid.is_csv_sheet(sheet):
+            # CSV 适配出来的「工作表」：从解析好的记录里取，口径同下（只给非空格、整行为空的不给）。分隔符用扫描时定下的
+            yield from csv_grid.rows(books.csv(csv_grid.sheet_delimiter(sheet)), min_row=min_row, max_row=max_row,
+                                     min_col=min_col, max_col=max_col)
             return
         yield from _sheet_rows(books, sheet, min_row=min_row, max_row=max_row, min_col=min_col, max_col=max_col)
     except UnsupportedTable:
diff --git a/backend/tests/test_csv_grid.py b/backend/tests/test_csv_grid.py
index 30a5f78a..dca02df7 100644
--- a/backend/tests/test_csv_grid.py
+++ b/backend/tests/test_csv_grid.py
@@ -8,7 +8,7 @@ import pytest
 
 from app.data import csv_grid, recipe_engine, xlsx_cells
 from app.data.recipe_types import Recipe
-from app.data.xlsx_scan import UnsupportedTable, scan_from_json, scan_to_json
+from app.data.xlsx_scan import SheetScan, UnsupportedTable, scan_from_json, scan_to_json
 
 CSV = ("订单号,下单日期,门店,金额,备注\n"
        "A001,2026-08-01,一号店,120,\n"
@@ -31,13 +31,13 @@ def test_scan_gives_one_visible_sheet_with_bounds_and_row_counts():
     scan = csv_grid.scan(CSV)
     assert scan.format == "csv" and len(scan.sheets) == 1
     sh = scan.sheets[0]
-    assert sh.name == csv_grid.CSV_SHEET and sh.state == "visible" and sh.path == csv_grid.CSV_PATH
+    assert sh.name == csv_grid.CSV_SHEET and sh.state == "visible" and sh.path == csv_grid.CSV_PATH + ","
     assert (sh.bounds.min_row, sh.bounds.min_col, sh.bounds.max_row, sh.bounds.max_col) == (1, 1, 5, 6)
     # 空字段不算非空格；第 4 行是空行
     assert sh.row_runs == [(1, 1, 5), (2, 2, 4), (3, 3, 5), (5, 5, 6)]
     assert sh.nonempty == 20
     again = scan_from_json(scan_to_json(scan))
-    assert again.format == "csv" and again.sheets[0].path == csv_grid.CSV_PATH
+    assert again.format == "csv" and again.sheets[0].path == csv_grid.CSV_PATH + ","
 
 
 def test_cell_values_are_typed_only_when_unambiguous():
@@ -71,9 +71,47 @@ def test_semicolon_and_tab_delimiters_and_gbk():
 
 
 def test_scan_any_dispatches_and_keeps_the_xlsx_refusal():
-    assert csv_grid.scan_any(CSV).format == "csv"
+    assert csv_grid.scan_any(CSV, "orders.csv").format == "csv"
+    assert csv_grid.scan_any(CSV, "ORDERS.CSV").format == "csv"
     with pytest.raises(UnsupportedTable):
-        csv_grid.scan_any(b"\xd0\xcf\x11\xe0\xa1\xb1\x1a\xe1" + b"\x00" * 100)
+        csv_grid.scan_any(b"\xd0\xcf\x11\xe0\xa1\xb1\x1a\xe1" + b"\x00" * 100, "old.xls")
+
+
+def test_only_csv_and_tsv_names_are_read_as_csv():
+    """复核反例：以前只看内容，不是 zip / OLE、前 64KB 没有 NUL 就当 CSV 读。系统导出的「HTML 表格冒充 .xls」、
+    误传的文本文件，都被拆成一张叫「CSV」的工作表去起草。现在和简单导入（tabular）一样按扩展名定：只有 .csv / .tsv
+    按 CSV 读；别的照旧交给 xlsx_scan 拒收。"""
+    with pytest.raises(UnsupportedTable) as e:
+        csv_grid.scan_any(CSV, "orders.txt")
+    assert "不是有效的 .xlsx" in str(e.value)
+    html = "<html><body><table><tr><td>订单号</td><td>金额</td></tr></table></body></html>".encode("utf-8")
+    for raw in (html, b"\xef\xbb\xbf  \r\n" + html, b"\xff\xfe" + "<table><tr><td>a</td></tr></table>".encode("utf-16-le"),
+                b'<?xml version="1.0"?><Workbook xmlns="urn:schemas-microsoft-com:office:spreadsheet"/>'):
+        with pytest.raises(UnsupportedTable) as e:
+            csv_grid.scan_any(raw, "report.xls")
+        assert "网页或 XML" in str(e.value) and "另存为 .xlsx 或 .csv" in str(e.value)
+    # 扩展名是 .csv、内容却不是文字（图片之类改了扩展名）：按 CSV 读不了，照实说
+    with pytest.raises(UnsupportedTable) as e:
+        csv_grid.scan_any(b"\x89PNG\r\n\x1a\n\x00\x00", "orders.csv")
+    assert "不是文字" in str(e.value)
+
+
+def test_tsv_is_always_tab_separated_and_the_delimiter_is_kept_for_reading():
+    """.tsv 一律按制表符分隔（同 tabular）：每行逗号个数一样的 TSV，嗅探会挑逗号（csv.Sniffer 优先逗号）。
+    扫描时定下的分隔符记在工作表的 path 里，之后读格子（重放、试运行）用同一个，不再各嗅各的。"""
+    raw = "名称,规格\t金额\n甲,1kg\t10\n乙,2kg\t20\n".encode("utf-8")      # 每行一个逗号、一个制表符
+    scan = csv_grid.scan_any(raw, "goods.tsv")
+    assert scan.sheets[0].path == csv_grid.CSV_PATH + "\t"
+    grid = xlsx_cells.read_grid(raw, scan_from_json(scan_to_json(scan)), csv_grid.CSV_SHEET)
+    assert grid.get(2, 1).value == "甲,1kg" and grid.get(2, 2).value == 10
+    # 这一版之前的暂存区：path 只有「csv:」，读的时候再嗅
+    assert csv_grid.is_csv_sheet(SheetScan(name="CSV", state="visible", path=csv_grid.CSV_PATH))
+    assert csv_grid.sheet_delimiter(SheetScan(name="CSV", state="visible", path=csv_grid.CSV_PATH)) is None
+    assert not csv_grid.is_csv_sheet(SheetScan(name="明细", state="visible", path="xl/worksheets/sheet1.xml"))
+    sniffed = csv_grid.scan_any(raw, "goods.csv")
+    assert sniffed.sheets[0].path == csv_grid.CSV_PATH + ","        # 同样的内容叫 .csv：照嗅探的来（逗号）
+    grid = xlsx_cells.read_grid(raw, sniffed, csv_grid.CSV_SHEET)
+    assert grid.get(2, 1).value == "甲" and grid.get(2, 2).value == "1kg\t10"
 
 
 def test_too_many_columns_is_refused():
diff --git a/backend/tests/test_recipe_imports.py b/backend/tests/test_recipe_imports.py
index cdfed879..74b54535 100644
--- a/backend/tests/test_recipe_imports.py
+++ b/backend/tests/test_recipe_imports.py
@@ -260,7 +260,7 @@ def fake_pipeline(state: FakeState) -> Pipeline:
 
     return Pipeline(
         validate=validate, recipe_sha256=recipe.recipe_sha256, canonical=recipe.canonical_recipe,
-        apply_patch=recipe.apply_patch, scan=lambda raw: fake_scan(raw, state),
+        apply_patch=recipe.apply_patch, scan=lambda raw, _filename: fake_scan(raw, state),
         scan_to_json=xlsx_scan.scan_to_json, scan_from_json=xlsx_scan.scan_from_json, read_grid=fake_grid,
         preview=xlsx_cells.preview, execute=execute, run_checks=run_checks, column_null_counts=null_counts,
         build_notes=build_notes, apply_notes=recipe_notes.apply_notes, detect_facts=detect_facts, draft=draft,
diff --git a/backend/tests/test_recipe_p5.py b/backend/tests/test_recipe_p5.py
index 9361fd42..163a7bf7 100644
--- a/backend/tests/test_recipe_p5.py
+++ b/backend/tests/test_recipe_p5.py
@@ -43,7 +43,7 @@ def _book(rows: list[tuple[Any, ...]], *, sheet: str = "明细", top: list[tuple
 
 
 def _run(raw: bytes, recipe: dict[str, Any], db: str | None = None, filename: str = "明细.xlsx", **kw: Any):
-    scan = csv_grid.scan_any(raw)
+    scan = csv_grid.scan_any(raw, filename)
     return recipe_engine.execute(Recipe.model_validate(recipe), raw, filename, scan, db, **kw)
 
 
@@ -314,7 +314,7 @@ def _ok_validate(data, facts, origin):
 
 
 def _draft(raw: bytes, filename: str = "明细.xlsx"):
-    scan = csv_grid.scan_any(raw)
+    scan = csv_grid.scan_any(raw, filename)
     grids = {s.name: xlsx_cells.read_grid(raw, scan, s.name, max_rows=recipe_suggest.DRAFT_MAX_ROWS)
              for s in scan.sheets if s.state == "visible" and s.nonempty}