diff --git a/picard/cluster.py b/picard/cluster.py
index 1cfac31c5..0399baaf7 100644
--- a/picard/cluster.py
+++ b/picard/cluster.py
@@ -46,6 +46,7 @@ from PyQt5 import QtCore
 
 from picard.config import get_config
 from picard.const import QUERY_LIMIT
+from picard.file import File
 from picard.metadata import (
     Metadata,
     SimMatchRelease,
@@ -317,7 +318,7 @@ class Cluster(FileList):
             if token:
                 cluster_list[token].add(album, artist or various_artists, file)
 
-        yield from (cluster for cluster in cluster_list.values() if len(cluster.files) > 1)
+        yield from cluster_list.values()
 
 
 class UnclusteredFiles(Cluster):
@@ -388,18 +389,22 @@ class ClusterList(list, Item):
 
 class FileCluster:
     def __init__(self):
-        self.files = []
+        self._files = []
         self._artist_counts = Counter()
         self._artists = defaultdict(Counter)
         self._titles = Counter()
 
     def add(self, album, artist, file):
-        self.files.append(file)
+        self._files.append(file)
         token = tokenize(artist)
         self._artist_counts[token] += 1
         self._artists[token][artist] += 1
         self._titles[album] += 1
 
+    @property
+    def files(self):
+        yield from (file for file in self._files if file.state != File.REMOVED)
+
     @property
     def artist(self):
         tokenized_artist = self._artist_counts.most_common(1)[0][0]
diff --git a/picard/tagger.py b/picard/tagger.py
index 6f3b7d226..3ad2a7edf 100644
--- a/picard/tagger.py
+++ b/picard/tagger.py
@@ -927,8 +927,10 @@ class Tagger(QtWidgets.QApplication):
         with self.window.ignore_selection_changes:
             self.window.set_sorting(False)
             for file_cluster in process_events_iter(result):
-                cluster = self.load_cluster(file_cluster.title, file_cluster.artist)
-                cluster.add_files(file_cluster.files)
+                files = set(file_cluster.files)
+                if len(files) > 1:
+                    cluster = self.load_cluster(file_cluster.title, file_cluster.artist)
+                    cluster.add_files(files)
             self.window.set_sorting(True)
 
         if callback:
diff --git a/test/test_clustering.py b/test/test_clustering.py
index cd86edd8d..ea3169ca3 100644
--- a/test/test_clustering.py
+++ b/test/test_clustering.py
@@ -73,8 +73,7 @@ class ClusterTest(PicardTestCase):
             self._create_file('album foo', 'artist foo'),
         ]
         clusters = list(Cluster.cluster(files))
-        # No cluster is being created for single files
-        self.assertEqual(0, len(clusters))
+        self.assertEqual(1, len(clusters))
 
     def test_cluster_single_cluster(self):
         files = [
@@ -92,12 +91,13 @@ class ClusterTest(PicardTestCase):
             self._create_file('album cluster2', 'artist foo'),
             self._create_file('album cluster1', 'artist foo'),
             self._create_file('albumcluster2', 'artist bar'),
-            self._create_file('album nocluster', 'artist bar'),
+            self._create_file('album single', 'artist bar'),
         ]
         clusters = list(Cluster.cluster(files))
-        self.assertEqual(2, len(clusters))
+        self.assertEqual(3, len(clusters))
         self.assertClusterEqual('album cluster1', 'artist bar', {files[0], files[2]}, clusters[0])
         self.assertClusterEqual('album cluster2', 'artist foo', {files[1], files[3]}, clusters[1])
+        self.assertClusterEqual('album single', 'artist bar', {files[4]}, clusters[2])
 
     def test_cluster_by_path(self):
         files = [
@@ -105,13 +105,14 @@ class ClusterTest(PicardTestCase):
             self._create_file(None, None, 'album2/foo1.ogg'),
             self._create_file(None, None, 'artist1/album1/foo2.ogg'),
             self._create_file(None, None, 'album2/foo2.ogg'),
-            self._create_file(None, None, 'nocluster/foo.ogg'),
+            self._create_file(None, None, 'single/foo.ogg'),
             self._create_file(None, None, 'album1/foo3.ogg'),
         ]
         clusters = list(Cluster.cluster(files))
-        self.assertEqual(2, len(clusters))
+        self.assertEqual(3, len(clusters))
         self.assertClusterEqual('album1', 'artist1', {files[0], files[2], files[5]}, clusters[0])
         self.assertClusterEqual('album2', 'Diverse Interpreten', {files[1], files[3]}, clusters[1])
+        self.assertClusterEqual('single', 'Diverse Interpreten', {files[4]}, clusters[2])
 
     def test_cluster_no_metadata(self):
         files = [
@@ -143,7 +144,7 @@ class FileClusterTest(PicardTestCase):
         fc.add('album 1', 'artist 1', file)
         self.assertEqual('album 1', fc.title)
         self.assertEqual('artist 1', fc.artist)
-        self.assertEqual([file], fc.files)
+        self.assertEqual([file], list(fc.files))
 
     def test_multi(self):
         files = [
@@ -161,4 +162,4 @@ class FileClusterTest(PicardTestCase):
         fc.add('album 2', 'Artist 1', files[4])
         self.assertEqual('Album 1', fc.title)
         self.assertEqual('Artist 1', fc.artist)
-        self.assertEqual(files, fc.files)
+        self.assertEqual(files, list(fc.files))