From c53260865e9c34b9d71b0f32be09f9c9278a2fe0 Mon Sep 17 00:00:00 2001
From: "Gareth.Y" <91742523+linjiaye929@users.noreply.github.com>
Date: Fri, 16 Oct 2026 04:03:05 -0700
Subject: [PATCH] fix: match standard metadata names case-insensitively

---
 crawl4ai/utils.py                     |  11 ++-
 tests/unit/test_metadata_name_case.py | 123 ++++++++++++++++++++++++++
 2 files changed, 131 insertions(+), 3 deletions(-)
 create mode 100644 tests/unit/test_metadata_name_case.py

diff --git a/crawl4ai/utils.py b/crawl4ai/utils.py
index b2e2cba7c..7468d946d 100644
--- a/crawl4ai/utils.py
+++ b/crawl4ai/utils.py
@@ -1584,16 +1584,21 @@ def extract_metadata_using_lxml(html, doc=None):
     # Strip and assign title
     metadata["title"] = title.strip() if title else None
 
+    # Standard metadata names are ASCII case-insensitive in HTML.
+    lowercase_name = (
+        'translate(@name, "ABCDEFGHIJKLMNOPQRSTUVWXYZ", "abcdefghijklmnopqrstuvwxyz")'
+    )
+
     # Meta description - using XPath with multiple attribute conditions
-    description = head.xpath('.//meta[@name="description"]/@content')
+    description = head.xpath(f'.//meta[{lowercase_name}="description"]/@content')
     metadata["description"] = description[0].strip() if description else None
 
     # Meta keywords
-    keywords = head.xpath('.//meta[@name="keywords"]/@content')
+    keywords = head.xpath(f'.//meta[{lowercase_name}="keywords"]/@content')
     metadata["keywords"] = keywords[0].strip() if keywords else None
 
     # Meta author
-    author = head.xpath('.//meta[@name="author"]/@content')
+    author = head.xpath(f'.//meta[{lowercase_name}="author"]/@content')
     metadata["author"] = author[0].strip() if author else None
 
     # Open Graph metadata - using starts-with() for performance
diff --git a/tests/unit/test_metadata_name_case.py b/tests/unit/test_metadata_name_case.py
new file mode 100644
index 000000000..a85dca628
--- /dev/null
+++ b/tests/unit/test_metadata_name_case.py
@@ -0,0 +1,123 @@
+"""Standard metadata names are ASCII case-insensitive in HTML."""
+
+import pytest
+
+from crawl4ai.content_scraping_strategy import LXMLWebScrapingStrategy
+
+
+def scrape(head):
+    html = (
+        "<html><head><title>Keep title</title>"
+        f"{head}</head><body><p>Meaningful article content remains here.</p>"
+        "</body></html>"
+    )
+    result = LXMLWebScrapingStrategy().scrap("https://example.test/article", html)
+    assert result.success
+    assert result.metadata["title"] == "Keep title"
+    return result.metadata
+
+
+@pytest.mark.parametrize("field", ["description", "keywords", "author"])
+@pytest.mark.parametrize("name_case", ["mixed", "upper", "lower"])
+def test_standard_metadata_names(field, name_case):
+    name = {"mixed": field.title(), "upper": field.upper(), "lower": field}[name_case]
+    content = "  Preserve MiXeD CASE and Café 中文  "
+    metadata = scrape(f'<meta name="{name}" content="{content}">')
+    assert metadata[field] == content.strip()
+
+
+def test_missing_metadata_fields():
+    metadata = scrape("")
+    assert {
+        field: metadata[field] for field in ["description", "keywords", "author"]
+    } == {"description": None, "keywords": None, "author": None}
+
+
+@pytest.mark.parametrize("name", ["Keywords", "keywordſ", "x-description"])
+def test_nonstandard_names_remain_unmatched(name):
+    metadata = scrape(f'<meta name="{name}" content="Must not match">')
+    assert metadata["description"] is None
+    assert metadata["keywords"] is None
+    assert metadata["author"] is None
+
+
+def test_duplicate_lowercase_names_keep_first_content():
+    metadata = scrape(
+        '<meta name="description" content="First VALUE">'
+        '<meta name="description" content="Second value">'
+    )
+    assert metadata["description"] == "First VALUE"
+
+
+def test_other_metadata_is_preserved():
+    metadata = scrape(
+        '<meta property="og:description" content="OG Summary">'
+        '<meta name="twitter:description" content="Twitter Summary">'
+        '<meta property="article:author" content="Article Author">'
+    )
+    assert metadata["og:description"] == "OG Summary"
+    assert metadata["twitter:description"] == "Twitter Summary"
+    assert metadata["article:author"] == "Article Author"
+    assert metadata["description"] is None
+    assert metadata["author"] is None
+
+
+@pytest.mark.parametrize("field", ["description", "keywords", "author"])
+@pytest.mark.parametrize("first_upper", [False, True])
+def test_duplicate_mixed_case_names_keep_first_content(field, first_upper):
+    first, second = (field.upper(), field) if first_upper else (field, field.upper())
+    metadata = scrape(
+        f'<meta name="{first}" content="First VALUE">'
+        f'<meta name="{second}" content="Second value">'
+    )
+    assert metadata[field] == "First VALUE"
+
+
+@pytest.mark.parametrize("field", ["description", "keywords", "author"])
+@pytest.mark.parametrize("name_case", ["lower", "upper"])
+def test_missing_content_is_skipped(field, name_case):
+    name = field if name_case == "lower" else field.upper()
+    metadata = scrape(
+        f'<meta name="{name}">' f'<meta name="{name}" content="Available VALUE">'
+    )
+    assert metadata[field] == "Available VALUE"
+
+
+@pytest.mark.parametrize("field", ["description", "keywords", "author"])
+def test_no_content_attribute_remains_none(field):
+    metadata = scrape(f'<meta name="{field}">')
+    assert metadata[field] is None
+
+
+@pytest.mark.parametrize("field", ["description", "keywords", "author"])
+def test_empty_content_remains_first_value(field):
+    metadata = scrape(
+        f'<meta name="{field}" content="">'
+        f'<meta name="{field}" content="Second value">'
+    )
+    assert metadata[field] == ""
+
+
+def test_social_names_keep_existing_case_matching():
+    metadata = scrape(
+        '<meta property="OG:description" content="Upper OG">'
+        '<meta name="Twitter:description" content="Mixed Twitter">'
+        '<meta property="Article:author" content="Mixed Article">'
+    )
+    assert "OG:description" not in metadata
+    assert "Twitter:description" not in metadata
+    assert "Article:author" not in metadata
+
+
+@pytest.mark.parametrize(
+    "head, expected",
+    [
+        ('<meta property="og:title" content="OpenGraph Title">', "OpenGraph Title"),
+        ('<meta name="twitter:title" content="Twitter Title">', "Twitter Title"),
+    ],
+)
+def test_existing_social_title_fallback_is_unchanged(head, expected):
+    html = f"<html><head>{head}</head><body><p>Article text is here.</p></body></html>"
+    result = LXMLWebScrapingStrategy().scrap("https://example.test/article", html)
+    assert result.success
+    assert result.metadata["title"] == expected