diff --git a/README.md b/README.md
index f4831e5..fd45c19 100644
--- a/README.md
+++ b/README.md
@@ -171,6 +171,15 @@ And this time, the specific configuration is applied.
 Caveats
 --------
 
+`pg_query_settings` doesn't work well with **prepared queries**. More specifically:
+for a given parameter `P`, if the generic plan is selected, and if the executor
+needs to fetch the value of `P`, then it won't get the value specified in the
+`pgqs_config` table. For example, the executor fetches the value of `work_mem`
+in order to decide whether to perform an in-memory sort, while the parameter
+`max_parallel_workers_per_gather` is only used during the planification stage,
+so `pg_query_settings` only works for this last example.
+This will be fixed in future releases.
+
 Enabling this extension might have a performance impact on any workloads
 with a high number of fast queries, as it slighlty increases the time taken
 to compute the query plan. The more entries in the table, the greater the