diff --git a/static/css/app.css b/static/css/app.css
index a2f0264..8242c99 100644
--- a/static/css/app.css
+++ b/static/css/app.css
@@ -1684,8 +1684,10 @@ body.dnd-mode .timer-display {
   border: 1px solid var(--border);
   border-radius: var(--radius);
   background: none;
+  color: var(--fg);
   text-align: left;
   cursor: pointer;
+  font: inherit;
 }
 .book-recent-item-title {
   font-size: 13px;
@@ -1717,8 +1719,10 @@ body.dnd-mode .timer-display {
   border: 1px solid var(--border);
   border-radius: var(--radius);
   background: none;
+  color: var(--fg);
   cursor: pointer;
   text-align: center;
+  font: inherit;
 }
 .book-folder-tile-icon {
   font-size: 22px;
diff --git a/static/js/app.js b/static/js/app.js
index dacfef4..30c3917 100644
--- a/static/js/app.js
+++ b/static/js/app.js
@@ -3618,6 +3618,35 @@ async function assignBookFolder(bookId) {
   }
 }
 
+// Fallback for books uploaded before ISBN extraction existed (or repaired/re-encrypted
+// without it): pulls the already-accessible encrypted book bytes, decrypts and unzips just
+// far enough to read the OPF identifiers — same source openBook() itself parses — without
+// doing a full reader render. EPUB only; PDFs never carried a reliable ISBN via pdf.js.
+async function _extractIsbnFromBookFile(bookId) {
+  try {
+    const key = await getOrCreateEncKey();
+    let data_ct, data_iv;
+    const cached = await _getCachedBook(bookId);
+    if (cached) {
+      ({data_ct, data_iv} = cached);
+    } else {
+      const res = await fetch(`/books/${bookId}/data/`);
+      ({data_ct, data_iv} = await res.json());
+    }
+    const plain = await decryptBytes(key, data_iv, data_ct);
+    const zip = await JSZip.loadAsync(plain);
+    const containerXml = await zip.file('META-INF/container.xml').async('text');
+    const opfPath = new DOMParser()
+      .parseFromString(containerXml, 'application/xml')
+      .querySelector('rootfile')?.getAttribute('full-path');
+    if (!opfPath) return '';
+    const opfDoc = new DOMParser().parseFromString(await zip.file(opfPath).async('text'), 'application/xml');
+    return _extractIsbnFromOpf(opfDoc);
+  } catch (e) {
+    return '';
+  }
+}
+
 // Manually triggered (never automatic) — looks up which library shelf/DDC category this
 // book falls under via the server-side DNB→Open Library proxy, purely as an organizational
 // hint (shown as a badge). See books/views.py:lookup_book_metadata for why this one call is
@@ -3625,20 +3654,34 @@ async function assignBookFolder(bookId) {
 async function lookupBookMetadata(bookId) {
   const book = _lastBookListData.find(b => b.id === bookId);
   if (!book) return;
-  const isbn = (bookMetaCache[bookId] || {}).isbn || book.isbn || '';
+  const cached = bookMetaCache[bookId] || {};
+  let isbn = cached.isbn || book.isbn || '';
+  const isPdf = (cached.type || book.type) === 'pdf';
+
   if (!isbn) {
-    await customAlert('Keine ISBN in den Metadaten dieses Buchs gefunden (nur bei neueren Uploads automatisch erkannt).');
-    return;
+    if (isPdf) {
+      await customAlert('Keine ISBN gespeichert — bei PDFs wird sie nicht automatisch erkannt.');
+      return;
+    }
+    // Older upload without a stored ISBN — extract it from the book file itself (may take
+    // a moment for large books, since it downloads/decrypts the full file if not cached).
+    isbn = await _extractIsbnFromBookFile(bookId);
+    if (!isbn) {
+      await customAlert('Keine ISBN in diesem Buch gefunden.');
+      return;
+    }
   }
+
   try {
     const res = await fetch(`/books/metadata-lookup/?isbn=${encodeURIComponent(isbn)}`);
     const data = await res.json();
-    if (!data.label) {
-      await customAlert('Keine Regal-Kategorie gefunden (weder bei der DNB noch bei Open Library).');
-      return;
-    }
-    await _updateBookMeta(bookId, {shelfTag: data.label});
+    // Persist the (possibly newly-found) ISBN either way, so a repeat lookup never needs
+    // to re-download/decrypt the book file again.
+    await _updateBookMeta(bookId, data.label ? {isbn, shelfTag: data.label} : {isbn});
     renderBookList(_lastBookListData);
+    if (!data.label) {
+      await customAlert('Keine Regal-Kategorie gefunden (weder bei der DNB noch bei Open Library). Die ISBN wurde für spätere Versuche gespeichert.');
+    }
   } catch (e) {
     await customAlert('Metadaten-Abruf fehlgeschlagen: ' + e.message);
   }
diff --git a/static/js/sw.js b/static/js/sw.js
index 5017b17..a683b55 100644
--- a/static/js/sw.js
+++ b/static/js/sw.js
@@ -2,7 +2,7 @@
  * diora service worker — caches the app shell for offline use.
  */
 
-const CACHE = 'diora-v37';
+const CACHE = 'diora-v38';
 const PODCAST_CACHE = 'diora-podcast-v1';
 const SHELL = [
   '/static/css/app.css',