diff --git a/pkg/labels/matcher.go b/pkg/labels/matcher.go
index 445c9056..f37fcb21 100644
--- a/pkg/labels/matcher.go
+++ b/pkg/labels/matcher.go
@@ -18,7 +18,9 @@ import (
 	"encoding/json"
 	"fmt"
 	"regexp"
+	"strconv"
 	"strings"
+	"unicode"
 
 	"github.com/prometheus/common/model"
 )
@@ -74,6 +76,9 @@ func NewMatcher(t MatchType, n, v string) (*Matcher, error) {
 }
 
 func (m *Matcher) String() string {
+	if strings.ContainsFunc(m.Name, isReserved) {
+		return fmt.Sprintf(`%s%s%s`, strconv.Quote(m.Name), m.Type, strconv.Quote(m.Value))
+	}
 	return fmt.Sprintf(`%s%s"%s"`, m.Name, m.Type, openMetricsEscape(m.Value))
 }
 
@@ -199,3 +204,13 @@ func (ms Matchers) String() string {
 
 	return buf.String()
 }
+
+// This is copied from matchers/parse/lexer.go. It will be removed when
+// the transition window from classic matchers to UTF-8 matchers is complete,
+// as then we can use double quotes when printing the label name for all
+// matchers. Until then, the classic parser does not understand double quotes
+// around the label name, so we use this function as a heuristic to tell if
+// the matcher was parsed with the UTF-8 parser or the classic parser.
+func isReserved(r rune) bool {
+	return unicode.IsSpace(r) || strings.ContainsRune("{}!=~,\\\"'`", r)
+}
diff --git a/pkg/labels/matcher_test.go b/pkg/labels/matcher_test.go
index 21d1b7f0..55d202f5 100644
--- a/pkg/labels/matcher_test.go
+++ b/pkg/labels/matcher_test.go
@@ -182,6 +182,30 @@ line`,
 			value: `tab	stop`,
 			want:  `foo="tab	stop"`,
 		},
+		{
+			name:  `foo`,
+			op:    MatchEqual,
+			value: `🙂`,
+			want:  `foo="🙂"`,
+		},
+		{
+			name:  `foo!`,
+			op:    MatchNotEqual,
+			value: `bar`,
+			want:  `"foo!"!="bar"`,
+		},
+		{
+			name:  `foo🙂`,
+			op:    MatchEqual,
+			value: `bar`,
+			want:  `foo🙂="bar"`,
+		},
+		{
+			name:  `foo bar`,
+			op:    MatchEqual,
+			value: `baz`,
+			want:  `"foo bar"="baz"`,
+		},
 	}
 
 	for _, test := range tests {