diff --git a/statement/src/_literal.rs b/statement/src/_literal.rs
index efe512a..4c46afb 100644
--- a/statement/src/_literal.rs
+++ b/statement/src/_literal.rs
@@ -36,7 +36,7 @@ pub enum Literal<'a> {
 
 impl Literal<'_> {
     /// Borrow this [`Literal`] as another [`Literal`].
-    pub fn borrowed(&self) -> Literal {
+    pub fn borrowed(&self) -> Literal<'_> {
         match self {
             Literal::Typed(lex, iri) => Literal::Typed(Cow::from(lex.as_ref()), iri.borrowed()),
             Literal::LanguageString(lex, lang_tag, base_dir) => {
@@ -46,7 +46,7 @@ impl Literal<'_> {
     }
 
     /// [lexical form](https://www.w3.org/TR/rdf12-concepts/#dfn-lexical-form) of this literal
-    pub fn lexical_form(&self) -> Cow<str> {
+    pub fn lexical_form(&self) -> Cow<'_, str> {
         let ref_cow = match self {
             Literal::Typed(lex, ..) => lex,
             Literal::LanguageString(lex, ..) => lex,
diff --git a/statement/src/_literal/_language_tag.rs b/statement/src/_literal/_language_tag.rs
index 94189d1..62120d2 100644
--- a/statement/src/_literal/_language_tag.rs
+++ b/statement/src/_literal/_language_tag.rs
@@ -1,4 +1,4 @@
-use std::borrow::Cow;
+use std::{borrow::Cow, cmp::Ordering};
 
 /// Wrapper around a [`Cow<str>`] guaranteeing that the underlying text satisfies [BCP47].
 ///
@@ -7,7 +7,7 @@ use std::borrow::Cow;
 /// (i.e. ISO 639 for 2-3 characters language tag, or ISO 15924 for the script)
 ///
 /// [BCP47]: https://datatracker.ietf.org/doc/bcp47/
-#[derive(Clone, Debug, Eq, Ord)]
+#[derive(Clone, Debug, Eq)]
 pub struct LangTag<'a>(Cow<'a, str>);
 
 impl<'a> LangTag<'a> {
@@ -76,25 +76,27 @@ impl std::cmp::PartialEq<LangTag<'_>> for &str {
     }
 }
 
+impl std::cmp::Ord for LangTag<'_> {
+    fn cmp(&self, other: &Self) -> std::cmp::Ordering {
+        cmp_ignore_case_ascii(&self.0, &other.0)
+    }
+}
+
 impl std::cmp::PartialOrd for LangTag<'_> {
     fn partial_cmp(&self, other: &Self) -> Option<std::cmp::Ordering> {
-        Some(
-            self.0
-                .to_ascii_lowercase()
-                .cmp(&other.0.to_ascii_lowercase()),
-        )
+        Some(self.cmp(other))
     }
 }
 
 impl std::cmp::PartialOrd<&str> for LangTag<'_> {
     fn partial_cmp(&self, other: &&'_ str) -> Option<std::cmp::Ordering> {
-        Some(self.0.to_ascii_lowercase().cmp(&other.to_ascii_lowercase()))
+        Some(cmp_ignore_case_ascii(&self.0, *other))
     }
 }
 
 impl std::cmp::PartialOrd<LangTag<'_>> for &str {
     fn partial_cmp(&self, other: &LangTag<'_>) -> Option<std::cmp::Ordering> {
-        Some(self.to_ascii_lowercase().cmp(&other.0.to_ascii_lowercase()))
+        Some(cmp_ignore_case_ascii(self, &other.0))
     }
 }
 
@@ -104,8 +106,26 @@ impl std::fmt::Display for LangTag<'_> {
     }
 }
 
+fn cmp_ignore_case_ascii(a: &str, b: &str) -> Ordering {
+    cmp_ignore_case_ascii_bytes(a.as_bytes(), b.as_bytes())
+}
+
+fn cmp_ignore_case_ascii_bytes(a: &[u8], b: &[u8]) -> Ordering {
+    match (a, b) {
+        ([], []) => Ordering::Equal,
+        ([], [_, ..]) => Ordering::Less,
+        ([_, ..], []) => Ordering::Greater,
+        ([ah, ar @ ..], [bh, br @ ..]) => ah
+            .to_ascii_lowercase()
+            .cmp(&bh.to_ascii_lowercase())
+            .then_with(|| cmp_ignore_case_ascii_bytes(ar, br)),
+    }
+}
+
 #[cfg(test)]
 mod test {
+    use std::cmp::Ordering;
+
     use super::*;
 
     #[test]
@@ -140,6 +160,10 @@ mod test {
         assert_eq!(tag1, tag2);
         assert_eq!(tag1, "en-gb");
         assert!(tag1 <= tag2 && tag2 <= tag1);
+        assert_eq!(tag1.partial_cmp(&tag2), Some(Ordering::Equal));
+        assert_eq!(tag2.partial_cmp(&tag1), Some(Ordering::Equal));
+        assert_eq!(tag1.cmp(&tag2), Ordering::Equal);
+        assert_eq!(tag2.cmp(&tag1), Ordering::Equal);
         assert!("EN" < tag1 && tag1 < "EN-ZZ");
     }
 }
diff --git a/statement/src/impl_oxrdf.rs b/statement/src/impl_oxrdf.rs
index c2cab47..b5d6a78 100644
--- a/statement/src/impl_oxrdf.rs
+++ b/statement/src/impl_oxrdf.rs
@@ -5,7 +5,7 @@
 //! This module is developed as if [`oxrdf`] implemented RDF 1.2 completely and strictly,
 //! which is not entirely true:
 //! - [`oxrdf`] does not support base direction in literals, so it is not complete;
-//! - [`oxrdf`] with the [`rdf-star`] feature allows triple terms in the subject position, so it is not strict.
+//! - [`oxrdf`] with the `rdf-star` feature allows triple terms in the subject position, so it is not strict.
 //!
 //! This is handled by panic'ing when those situations are encountered.
 //!