diff --git a/src/__phutil_library_map__.php b/src/__phutil_library_map__.php
--- a/src/__phutil_library_map__.php
+++ b/src/__phutil_library_map__.php
@@ -561,6 +561,7 @@
     'phutil_utf8_encode_codepoint' => 'utils/utf8.php',
     'phutil_utf8_hard_wrap' => 'utils/utf8.php',
     'phutil_utf8_hard_wrap_html' => 'utils/utf8.php',
+    'phutil_utf8_is_cjk' => 'utils/utf8.php',
     'phutil_utf8_is_combining_character' => 'utils/utf8.php',
     'phutil_utf8_strlen' => 'utils/utf8.php',
     'phutil_utf8_strtolower' => 'utils/utf8.php',
diff --git a/src/search/PhutilSearchQueryCompiler.php b/src/search/PhutilSearchQueryCompiler.php
--- a/src/search/PhutilSearchQueryCompiler.php
+++ b/src/search/PhutilSearchQueryCompiler.php
@@ -271,6 +271,8 @@
         continue;
       }
 
+      $is_quoted = $token['quoted'];
+
       switch ($operator_string) {
         case '-':
           $operator = self::OPERATOR_NOT;
@@ -281,10 +283,20 @@
         case '=':
           $operator = self::OPERATOR_EXACT;
           break;
-        case '':
         case '+':
           $operator = self::OPERATOR_AND;
           break;
+        case '':
+          // See T12995. If this query term contains Chinese, Japanese or
+          // Korean characters, treat the term as a substring term by default.
+          // These languages do not separate words with spaces, so the term
+          // search mode is normally useless.
+          if ($enable_functions && !$is_quoted && phutil_utf8_is_cjk($value)) {
+            $operator = self::OPERATOR_SUBSTRING;
+          } else {
+            $operator = self::OPERATOR_AND;
+          }
+          break;
         default:
           throw new PhutilSearchQueryCompilerSyntaxException(
             pht(
@@ -294,7 +306,7 @@
 
       $result = array(
         'operator' => $operator,
-        'quoted' => $token['quoted'],
+        'quoted' => $is_quoted,
         'value' => $value,
       );
 
diff --git a/src/search/__tests__/PhutilSearchQueryCompilerTestCase.php b/src/search/__tests__/PhutilSearchQueryCompilerTestCase.php
--- a/src/search/__tests__/PhutilSearchQueryCompilerTestCase.php
+++ b/src/search/__tests__/PhutilSearchQueryCompilerTestCase.php
@@ -92,6 +92,8 @@
     $op_sub = PhutilSearchQueryCompiler::OPERATOR_SUBSTRING;
     $op_exact = PhutilSearchQueryCompiler::OPERATOR_EXACT;
 
+    $mao = "\xE7\x8C\xAB";
+
     $function_tests = array(
       'cat' => array(
         array(null, $op_and, 'cat'),
@@ -119,6 +121,18 @@
       '~"core and seven years ag"' => array(
         array(null, $op_sub, 'core and seven years ag'),
       ),
+      $mao => array(
+        array(null, $op_sub, $mao),
+      ),
+      '+'.$mao => array(
+        array(null, $op_and, $mao),
+      ),
+      '~'.$mao => array(
+        array(null, $op_sub, $mao),
+      ),
+      '"'.$mao.'"' => array(
+        array(null, $op_and, $mao),
+      ),
     );
 
     $this->assertCompileFunctionQueries($function_tests);
diff --git a/src/utils/__tests__/PhutilUTF8TestCase.php b/src/utils/__tests__/PhutilUTF8TestCase.php
--- a/src/utils/__tests__/PhutilUTF8TestCase.php
+++ b/src/utils/__tests__/PhutilUTF8TestCase.php
@@ -609,6 +609,22 @@
     $this->assertTrue(true);
   }
 
+  public function testCJK() {
+    $map = array(
+      '' => false,
+      'a' => false,
+      '.' => false,
+      "\xE2\x98\x83" => false,
+      "\xE5\xA0\xB1" => true,
+    );
+
+    foreach ($map as $input => $expect) {
+      $actual = phutil_utf8_is_cjk($input);
+
+      $this->assertEqual($expect, $actual, pht('CJK: "%s"', $input));
+    }
+  }
+
   public function testUTF8BMP() {
     $tests = array(
       ''  => array(
diff --git a/src/utils/utf8.php b/src/utils/utf8.php
--- a/src/utils/utf8.php
+++ b/src/utils/utf8.php
@@ -327,6 +327,64 @@
 }
 
 
+/**
+ * Test if a string contains Chinese, Japanese, or Korean characters.
+ *
+ * Most languages use spaces to separate words, but these languages do not.
+ *
+ * @param string String to examine, in UTF8.
+ * @return bool True if the string contains Chinese, Japanese, or Korean
+ *   characters.
+ */
+function phutil_utf8_is_cjk($string) {
+  $codepoints = phutil_utf8v_codepoints($string);
+
+  foreach ($codepoints as $codepoint) {
+    // CJK Unified Ideographs
+    if ($codepoint >= 0x4E00 && $codepoint <= 0x9FFF) {
+      return true;
+    }
+
+    // CJK Unified Ideographs Extension A
+    if ($codepoint >= 0x3400 && $codepoint <= 0x4DBF) {
+      return true;
+    }
+
+    // CJK Unified Ideographs Extension B
+    if ($codepoint >= 0x20000 && $codepoint <= 0x2A6DF) {
+      return true;
+    }
+
+    // CJK Unified Ideographs Extension C
+    if ($codepoint >= 0x2A700 && $codepoint <= 0x2B73F) {
+      return true;
+    }
+
+    // CJK Unified Ideographs Extension D
+    if ($codepoint >= 0x2B740 && $codepoint <= 0x2B81F) {
+      return true;
+    }
+
+    // CJK Unified Ideographs Extension E
+    if ($codepoint >= 0x2B820 && $codepoint <= 0x2CEAF) {
+      return true;
+    }
+
+    // CJK Unified Ideographs Extension F
+    if ($codepoint >= 0x2CEB0 && $codepoint <= 0x2EBEF) {
+      return true;
+    }
+
+    // CJK Compatibility Ideographs
+    if ($codepoint >= 0xF900 && $codepoint <= 0xFAFF) {
+      return true;
+    }
+  }
+
+  return false;
+}
+
+
 /**
  * Split a UTF-8 string into an array of characters. Combining characters are
  * also split.