diff --git a/components/esp_http_server/src/httpd_parse.c b/components/esp_http_server/src/httpd_parse.c
index b520c871b1..62244f9f9f 100644
--- a/components/esp_http_server/src/httpd_parse.c
+++ b/components/esp_http_server/src/httpd_parse.c
@@ -774,9 +774,16 @@ esp_err_t httpd_req_new(struct httpd_data *hd, struct sock_db *sd)
             ESP_LOGD(TAG, LOG_FMT("Received PONG frame"));
         }
 
-        /* Call handler if it's a non-control frame (or if handler requests control frames, as well) */
+        /* Call handler if it's a non-control frame, a PONG frame,
+         * or if handler requests control frames as well.
+         * PONG must be dispatched so that:
+         *  1. User code that sends PINGs can track responses (heartbeat)
+         *  2. The PONG frame bytes are consumed from the socket via
+         *     httpd_ws_recv_frame(), preventing TCP stream misalignment */
         if (ret == ESP_OK &&
-            (ra->ws_type < HTTPD_WS_TYPE_CLOSE || sd->ws_control_frames)) {
+            (ra->ws_type < HTTPD_WS_TYPE_CLOSE ||
+             ra->ws_type == HTTPD_WS_TYPE_PONG ||
+             sd->ws_control_frames)) {
             ret = sd->ws_handler(r);
         }