diff --git a/Tests/LibWeb/TestHTMLTokenizer.cpp b/Tests/LibWeb/TestHTMLTokenizer.cpp
index 36f2cec007c..7c844696187 100644
--- a/Tests/LibWeb/TestHTMLTokenizer.cpp
+++ b/Tests/LibWeb/TestHTMLTokenizer.cpp
@@ -57,7 +57,7 @@ using Token = Web::HTML::HTMLToken;
 
 #define EXPECT_TAG_TOKEN_ATTRIBUTE(name, value) \
     VERIFY(last_token);                         \
-    EXPECT_EQ(last_token->attribute(#name), #value);
+    EXPECT_EQ(last_token->attribute(#name), value);
 
 #define EXPECT_TAG_TOKEN_ATTRIBUTE_COUNT(count) \
     VERIFY(last_token);                         \
@@ -124,7 +124,7 @@ TEST_CASE(unquoted_attributes)
     BEGIN_ENUMERATION(tokens);
     EXPECT_START_TAG_TOKEN(p);
     EXPECT_TAG_TOKEN_ATTRIBUTE_COUNT(1);
-    EXPECT_TAG_TOKEN_ATTRIBUTE(foo, bar);
+    EXPECT_TAG_TOKEN_ATTRIBUTE(foo, "bar");
     EXPECT_END_OF_FILE_TOKEN();
     END_ENUMERATION();
 }
@@ -135,7 +135,7 @@ TEST_CASE(single_quoted_attributes)
     BEGIN_ENUMERATION(tokens);
     EXPECT_START_TAG_TOKEN(p);
     EXPECT_TAG_TOKEN_ATTRIBUTE_COUNT(1);
-    EXPECT_TAG_TOKEN_ATTRIBUTE(foo, bar);
+    EXPECT_TAG_TOKEN_ATTRIBUTE(foo, "bar");
     EXPECT_END_OF_FILE_TOKEN();
     END_ENUMERATION();
 }
@@ -146,7 +146,7 @@ TEST_CASE(double_quoted_attributes)
     BEGIN_ENUMERATION(tokens);
     EXPECT_START_TAG_TOKEN(p);
     EXPECT_TAG_TOKEN_ATTRIBUTE_COUNT(1);
-    EXPECT_TAG_TOKEN_ATTRIBUTE(foo, bar);
+    EXPECT_TAG_TOKEN_ATTRIBUTE(foo, "bar");
     EXPECT_END_OF_FILE_TOKEN();
     END_ENUMERATION();
 }
@@ -157,9 +157,22 @@ TEST_CASE(multiple_attributes)
     BEGIN_ENUMERATION(tokens);
     EXPECT_START_TAG_TOKEN(p);
     EXPECT_TAG_TOKEN_ATTRIBUTE_COUNT(3);
-    EXPECT_TAG_TOKEN_ATTRIBUTE(foo, bar);
-    EXPECT_TAG_TOKEN_ATTRIBUTE(baz, foobar);
-    EXPECT_TAG_TOKEN_ATTRIBUTE(foo2, bar2);
+    EXPECT_TAG_TOKEN_ATTRIBUTE(foo, "bar");
+    EXPECT_TAG_TOKEN_ATTRIBUTE(baz, "foobar");
+    EXPECT_TAG_TOKEN_ATTRIBUTE(foo2, "bar2");
+    EXPECT_END_OF_FILE_TOKEN();
+    END_ENUMERATION();
+}
+
+TEST_CASE(character_reference_in_attribute)
+{
+    auto tokens = run_tokenizer("<p foo=a&amp;b bar='a&amp;b' baz=\"a&amp;b\">");
+    BEGIN_ENUMERATION(tokens);
+    EXPECT_START_TAG_TOKEN(p);
+    EXPECT_TAG_TOKEN_ATTRIBUTE_COUNT(3);
+    EXPECT_TAG_TOKEN_ATTRIBUTE(foo, "a&b");
+    EXPECT_TAG_TOKEN_ATTRIBUTE(bar, "a&b");
+    EXPECT_TAG_TOKEN_ATTRIBUTE(baz, "a&b");
     EXPECT_END_OF_FILE_TOKEN();
     END_ENUMERATION();
 }
diff --git a/Userland/Libraries/LibWeb/HTML/Parser/HTMLTokenizer.cpp b/Userland/Libraries/LibWeb/HTML/Parser/HTMLTokenizer.cpp
index aa6efbe72ee..5fc69491e04 100644
--- a/Userland/Libraries/LibWeb/HTML/Parser/HTMLTokenizer.cpp
+++ b/Userland/Libraries/LibWeb/HTML/Parser/HTMLTokenizer.cpp
@@ -1179,9 +1179,8 @@ _StartOfFunction:
                 }
                 ON('&')
                 {
-                    m_current_token.last_attribute().value = consume_current_builder();
                     m_return_state = State::AttributeValueDoubleQuoted;
-                    SWITCH_TO(CharacterReference);
+                    SWITCH_TO_WITH_UNCLEAN_BUILDER(CharacterReference);
                 }
                 ON(0)
                 {
@@ -1211,9 +1210,8 @@ _StartOfFunction:
                 }
                 ON('&')
                 {
-                    m_current_token.last_attribute().value = consume_current_builder();
                     m_return_state = State::AttributeValueSingleQuoted;
-                    SWITCH_TO(CharacterReference);
+                    SWITCH_TO_WITH_UNCLEAN_BUILDER(CharacterReference);
                 }
                 ON(0)
                 {
@@ -1244,9 +1242,8 @@ _StartOfFunction:
                 }
                 ON('&')
                 {
-                    m_current_token.last_attribute().value = consume_current_builder();
                     m_return_state = State::AttributeValueUnquoted;
-                    SWITCH_TO(CharacterReference);
+                    SWITCH_TO_WITH_UNCLEAN_BUILDER(CharacterReference);
                 }
                 ON('>')
                 {