diff --git a/src/cmd/frequency.rs b/src/cmd/frequency.rs
index 2cef402db..66cf2f4a4 100644
--- a/src/cmd/frequency.rs
+++ b/src/cmd/frequency.rs
@@ -208,7 +208,7 @@ impl Args {
                 for (i, field) in nsel.select(row_work.into_iter()).enumerate() {
                     field_work = {
                         if let Ok(s) = simdutf8::basic::from_utf8(field) {
-                            to_lowercase_into(s.trim(), &mut buf);
+                            util::to_lowercase_into(s.trim(), &mut buf);
                             buf.as_bytes().to_vec()
                         } else {
                             field.to_vec()
@@ -252,17 +252,3 @@ impl Args {
         Ok((sel.select(headers).map(<[u8]>::to_vec).collect(), sel))
     }
 }
-
-/// this is a non-allocating to_lowercase that uses an existing buffer
-/// and should be faster than the stdlib version
-/// TODO: if this proves to be faster per the benchmarks, we should use
-/// this project-wide over the allocating stdlib version, and move it to utils.rs
-#[inline]
-fn to_lowercase_into(s: &str, buf: &mut String) {
-    buf.clear();
-    for c in s.chars() {
-        for lc in c.to_lowercase() {
-            buf.push(lc);
-        }
-    }
-}
diff --git a/src/util.rs b/src/util.rs
index dc5cee522..5a819c491 100644
--- a/src/util.rs
+++ b/src/util.rs
@@ -1407,6 +1407,18 @@ pub async fn download_file(
     Ok(())
 }
 
+/// this is a non-allocating to_lowercase that uses an existing buffer
+/// and should be faster than the allocating std::to_lowercase
+#[inline]
+pub fn to_lowercase_into(s: &str, buf: &mut String) {
+    buf.clear();
+    for c in s.chars() {
+        for lc in c.to_lowercase() {
+            buf.push(lc);
+        }
+    }
+}
+
 /// load the first BUFFER*4 (128k) bytes of the file and check if it is utf8
 pub fn isutf8_file(path: &Path) -> Result<bool, CliError> {
     let metadata = std::fs::metadata(path)?;