diff --git a/internal/frontend/goldmark.go b/internal/frontend/goldmark.go
index ab75979..4ba5666 100644
--- a/internal/frontend/goldmark.go
+++ b/internal/frontend/goldmark.go
@@ -12,7 +12,6 @@
 	"fmt"
 	"regexp"
 	"strings"
-	"unicode"
 
 	"github.com/yuin/goldmark/ast"
 	"github.com/yuin/goldmark/parser"
@@ -180,7 +179,7 @@
 }
 
 // Generate turns heading content from a markdown document into a heading id.
-// First HTML markup and markdown images are stripped then unicode letters
+// First HTML markup and markdown images are stripped then ASCII letters
 // and numbers are used to generate the final result. Finally, all heading ids
 // are prefixed with "readme-" to avoid name collisions with other ids on the
 // unit page. Duplicated heading ids are given an incremental suffix. See
@@ -190,7 +189,7 @@
 	r := regexp.MustCompile(`(<[^<>]+>|\[\!\[[^\]]+]\([^\)]+\)\]\([^\)]+\))`)
 	str := r.ReplaceAllString(string(value), "")
 	f := func(c rune) bool {
-		return !unicode.IsLetter(c) && !unicode.IsNumber(c)
+		return !('a' <= c && c <= 'z') && !('A' <= c && c <= 'Z') && !('0' <= c && c <= '9')
 	}
 	str = strings.Join(strings.FieldsFunc(str, f), "-")
 	str = strings.ToLower(str)
diff --git a/internal/frontend/readme_test.go b/internal/frontend/readme_test.go
index 9878452..be41a09 100644
--- a/internal/frontend/readme_test.go
+++ b/internal/frontend/readme_test.go
@@ -145,6 +145,16 @@
 			},
 		},
 		{
+			name: "Non-ASCII Heading",
+			unit: unit,
+			readme: &internal.Readme{
+				Filepath: sample.ReadmeFilePath,
+				Contents: "# 中文¹",
+			},
+			wantHTML:    "<h3 class=\"h1\" id=\"readme-heading\">中文¹</h3>",
+			wantOutline: []*Heading{{Level: 1, Text: "中文¹", ID: "readme-heading"}},
+		},
+		{
 			name: "Github markdown emoji markup is properly rendered",
 			unit: unit,
 			readme: &internal.Readme{