diff --git a/keras/src/applications/densenet.py b/keras/src/applications/densenet.py
index 5e0247715053..886b6bc16bd6 100644
--- a/keras/src/applications/densenet.py
+++ b/keras/src/applications/densenet.py
@@ -170,7 +170,7 @@ def DenseNet(
                 be applied.
         classes: optional number of classes to classify images
             into, only to be specified if `include_top` is `True`, and
-            if no `weights` argument is specified.
+            if no `weights` argument is specified. Defaults to `1000`.
         classifier_activation: A `str` or callable.
             The activation function to use
             on the "top" layer. Ignored unless `include_top=True`. Set
diff --git a/keras/src/applications/resnet.py b/keras/src/applications/resnet.py
index 5e735a9ee1b7..0948f8901db1 100644
--- a/keras/src/applications/resnet.py
+++ b/keras/src/applications/resnet.py
@@ -574,7 +574,7 @@ def decode_predictions(preds, top=5):
         - `max` means that global max pooling will be applied.
     classes: optional number of classes to classify images into, only to be
         specified if `include_top` is `True`, and if no `weights` argument is
-        specified.
+        specified. Defaults to `1000`.
     classifier_activation: A `str` or callable. The activation function to
         use on the "top" layer. Ignored unless `include_top=True`. Set
         `classifier_activation=None` to return the logits of the "top" layer.
diff --git a/keras/src/applications/xception.py b/keras/src/applications/xception.py
index 48eda59f4fcf..2464d45ae2a2 100644
--- a/keras/src/applications/xception.py
+++ b/keras/src/applications/xception.py
@@ -81,7 +81,7 @@ def Xception(
                 be applied.
         classes: optional number of classes to classify images
             into, only to be specified if `include_top` is `True`, and
-            if no `weights` argument is specified.
+            if no `weights` argument is specified. Defaults to `1000`.
         classifier_activation: A `str` or callable. The activation function to
             use on the "top" layer. Ignored unless `include_top=True`. Set
             `classifier_activation=None` to return the logits of the "top"