diff --git a/sql/core/src/main/scala/org/apache/spark/sql/Dataset.scala b/sql/core/src/main/scala/org/apache/spark/sql/Dataset.scala
index 7be4aa1ca9562..06ff601d67a1d 100644
--- a/sql/core/src/main/scala/org/apache/spark/sql/Dataset.scala
+++ b/sql/core/src/main/scala/org/apache/spark/sql/Dataset.scala
@@ -990,7 +990,7 @@ class Dataset[T] private[sql](
    * @param condition Join expression.
    * @param joinType Type of join to perform. Default `inner`. Must be one of:
    *                 `inner`, `cross`, `outer`, `full`, `full_outer`, `left`, `left_outer`,
-   *                 `right`, `right_outer`, `left_semi`, `left_anti`.
+   *                 `right`, `right_outer`.
    *
    * @group typedrel
    * @since 1.6.0
@@ -1007,6 +1007,10 @@ class Dataset[T] private[sql](
         JoinType(joinType),
         Some(condition.expr))).analyzed.asInstanceOf[Join]
 
+    if (joined.joinType == LeftSemi || joined.joinType == LeftAnti) {
+      throw new AnalysisException("Invalid join type in joinWith: " + joined.joinType.sql)
+    }
+
     // For both join side, combine all outputs into a single column and alias it with "_1" or "_2",
     // to match the schema for the encoder of the join result.
     // Note that we do this before joining them, to enable the join operator to return null for one
diff --git a/sql/core/src/test/scala/org/apache/spark/sql/DatasetSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/DatasetSuite.scala
index 87b7b090de3bf..4a1e3e1599599 100644
--- a/sql/core/src/test/scala/org/apache/spark/sql/DatasetSuite.scala
+++ b/sql/core/src/test/scala/org/apache/spark/sql/DatasetSuite.scala
@@ -21,6 +21,7 @@ import java.io.{Externalizable, ObjectInput, ObjectOutput}
 import java.sql.{Date, Timestamp}
 
 import org.apache.spark.sql.catalyst.encoders.{OuterScopes, RowEncoder}
+import org.apache.spark.sql.catalyst.plans.{LeftAnti, LeftSemi}
 import org.apache.spark.sql.catalyst.util.sideBySide
 import org.apache.spark.sql.execution.{LogicalRDD, RDDScanExec, SortExec}
 import org.apache.spark.sql.execution.exchange.{BroadcastExchangeExec, ShuffleExchange}
@@ -321,6 +322,21 @@ class DatasetSuite extends QueryTest with SharedSQLContext {
       ((("b", 2), ("b", 2)), ("b", 2)))
   }
 
+  test("joinWith join types") {
+    val ds1 = Seq(1, 2, 3).toDS().as("a")
+    val ds2 = Seq(1, 2).toDS().as("b")
+
+    val e1 = intercept[AnalysisException] {
+      ds1.joinWith(ds2, $"a.value" === $"b.value", "left_semi")
+    }.getMessage
+    assert(e1.contains("Invalid join type in joinWith: " + LeftSemi.sql))
+
+    val e2 = intercept[AnalysisException] {
+      ds1.joinWith(ds2, $"a.value" === $"b.value", "left_anti")
+    }.getMessage
+    assert(e2.contains("Invalid join type in joinWith: " + LeftAnti.sql))
+  }
+
   test("groupBy function, keys") {
     val ds = Seq(("a", 1), ("b", 1)).toDS()
     val grouped = ds.groupByKey(v => (1, v._2))