diff --git a/rslib/src/media/check.rs b/rslib/src/media/check.rs
index c901583da..d7adfa361 100644
--- a/rslib/src/media/check.rs
+++ b/rslib/src/media/check.rs
@@ -723,9 +723,19 @@ Unused: unused.jpg
 
     #[test]
     fn html_encoding() {
-        let field = "[sound:a & b.mp3]";
+        let mut field = "[sound:a & b.mp3]";
         let mut seen = Default::default();
         normalize_and_maybe_rename_files(field, &HashMap::new(), &mut seen, Path::new("/tmp"));
         assert!(seen.contains("a & b.mp3"));
+
+        field = r#"<img src="a&b.jpg">"#;
+        seen = Default::default();
+        normalize_and_maybe_rename_files(field, &HashMap::new(), &mut seen, Path::new("/tmp"));
+        assert!(seen.contains("a&b.jpg"));
+
+        field = r#"<img src="a&amp;b.jpg">"#;
+        seen = Default::default();
+        normalize_and_maybe_rename_files(field, &HashMap::new(), &mut seen, Path::new("/tmp"));
+        assert!(seen.contains("a&b.jpg"));
     }
 }
diff --git a/rslib/src/text.rs b/rslib/src/text.rs
index bb4176861..26773b436 100644
--- a/rslib/src/text.rs
+++ b/rslib/src/text.rs
@@ -219,7 +219,7 @@ pub(crate) fn extract_media_refs(text: &str) -> Vec<MediaRef> {
             .or_else(|| caps.get(3))
             .unwrap()
             .as_str();
-        let fname_decoded = fname.into();
+        let fname_decoded = decode_entities(fname);
         out.push(MediaRef {
             full_ref: caps.get(0).unwrap().as_str(),
             fname,