diff --git a/common/data_source/html_utils.py b/common/data_source/html_utils.py
index b39569e664..9fd33a7b0e 100644
--- a/common/data_source/html_utils.py
+++ b/common/data_source/html_utils.py
@@ -76,12 +76,24 @@ def format_document_soup(document: bs4.BeautifulSoup, table_cell_separator: str
     text = ""
     list_element_start = False
     verbatim_output = 0
-    in_table = False
     last_added_newline = False
-    link_href: str | None = None
+
+    # ``descendants`` yields opening tags only, so a flag set on <table>/<a> would
+    # never clear. Precompute scope by registering each <table>/<a> descendant;
+    # this avoids O(depth) ancestor walks while keeping lookups O(1).
+    table_scope = {id(d) for table in document.find_all("table") for d in table.descendants}
+    href_scope = {}
+    for anchor in document.find_all("a"):  # document order, so a nested <a> wins over its parent
+        href_value = anchor.get("href", None)
+        # mostly for typing, having multiple hrefs is not valid HTML
+        link_href = href_value[0] if isinstance(href_value, list) else href_value
+        href_scope.update((id(d), link_href) for d in anchor.descendants)
 
     for e in document.descendants:
         verbatim_output -= 1
+        in_table = id(e) in table_scope
+        link_href = href_scope.get(id(e))
+
         if isinstance(e, bs4.element.NavigableString):
             if isinstance(e, (bs4.element.Comment, bs4.element.Doctype)):
                 continue
@@ -109,26 +121,15 @@ def format_document_soup(document: bs4.BeautifulSoup, table_cell_separator: str
 
                 list_element_start = False
         elif isinstance(e, bs4.element.Tag):
-            # table is standard HTML element
-            if e.name == "table":
-                in_table = True
             # TR is for rows
-            elif e.name == "tr" and in_table:
+            if e.name == "tr" and in_table:
                 text += "\n"
             # td for data cell, th for header
             elif e.name in ["td", "th"] and in_table:
                 text += table_cell_separator
-            elif e.name == "/table":
-                in_table = False
             elif in_table:
                 # don't handle other cases while in table
                 pass
-            elif e.name == "a":
-                href_value = e.get("href", None)
-                # mostly for typing, having multiple hrefs is not valid HTML
-                link_href = href_value[0] if isinstance(href_value, list) else href_value
-            elif e.name == "/a":
-                link_href = None
             elif e.name in ["p", "div"]:
                 if not list_element_start:
                     text += "\n"
diff --git a/test/unit_test/data_source/test_html_utils.py b/test/unit_test/data_source/test_html_utils.py
new file mode 100644
index 0000000000..242ae90d3b
--- /dev/null
+++ b/test/unit_test/data_source/test_html_utils.py
@@ -0,0 +1,91 @@
+#
+#  Copyright 2025 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+import bs4
+import pytest
+
+from common.data_source import html_utils
+from common.data_source.config import HtmlBasedConnectorTransformLinksStrategy
+from common.data_source.html_utils import format_document_soup
+
+
+def _fmt(html: str) -> str:
+    return format_document_soup(bs4.BeautifulSoup(html, "html.parser"))
+
+
+@pytest.fixture
+def markdown_links(monkeypatch):
+    """``format_element_text`` only renders links under the markdown strategy."""
+    monkeypatch.setattr(
+        html_utils,
+        "HTML_BASED_CONNECTOR_TRANSFORM_LINKS_STRATEGY",
+        HtmlBasedConnectorTransformLinksStrategy.MARKDOWN,
+    )
+
+
+TABLE = "<table><tr><td>A</td><td>B</td></tr></table>"
+
+
+def test_paragraph_after_table_keeps_its_newline():
+    assert "\nAfter" in _fmt(f"<p>Before</p>{TABLE}<p>After</p>")
+
+
+def test_list_after_table_keeps_hyphen_markers():
+    assert "\n- item1" in _fmt(f"<p>Before</p>{TABLE}<ul><li>item1</li><li>item2</li></ul>")
+
+
+def test_block_after_table_starts_on_a_new_line():
+    # The div must not be folded onto the last table row.
+    assert "\nTrailing" in _fmt(f"{TABLE}<div>Trailing</div>")
+
+
+def test_table_still_separates_rows_and_cells():
+    # Control: the table itself must keep working — rows on newlines, cells tab-separated.
+    assert _fmt("<table><tr><td>A</td><td>B</td></tr><tr><td>C</td><td>D</td></tr></table>") == "A\tB\n\tC\tD"
+
+
+def test_content_after_table_matches_the_same_content_without_a_table(markdown_links):
+    tail = "<p>After</p><ul><li>item1</li></ul>"
+    with_table = _fmt(f"{TABLE}{tail}")
+    without_table = _fmt(tail)
+    assert with_table.endswith(without_table.lstrip("\n"))
+
+
+def test_text_after_link_in_same_paragraph_is_not_linkified(markdown_links):
+    assert _fmt('<p>see <a href="http://x.com">link</a> after</p>') == "see [link](http://x.com) after"
+
+
+def test_paragraph_after_link_is_not_linkified(markdown_links):
+    out = _fmt('<p><a href="http://x.com">link</a></p><p>next paragraph</p>')
+    assert "[next paragraph]" not in out
+
+
+def test_link_inside_anchor_is_still_linkified(markdown_links):
+    # The anchor text itself must keep its markdown link.
+    assert _fmt('<p><a href="http://x.com">link</a></p>') == "[link](http://x.com)"
+
+
+def test_table_cells_do_not_inherit_a_preceding_links_href(markdown_links):
+    out = _fmt(f'<p><a href="http://x.com">pre</a></p>{TABLE}')
+    assert "[A](http://x.com)" not in out
+
+
+def test_link_inside_a_table_cell_is_linkified(markdown_links):
+    assert _fmt('<table><tr><td><a href="http://x.com">cell</a></td></tr></table>') == "[cell](http://x.com)"
+
+
+def test_link_is_stripped_under_the_default_strategy():
+    # Default strategy is STRIP: no markdown link syntax at all.
+    assert _fmt('<p>see <a href="http://x.com">link</a> after</p>') == "see link after"