diff --git a/NEWS.md b/NEWS.md
index 4c598039..ba4703be 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -1,5 +1,7 @@
 # mlr3tuning (development version)
 
+* fix: The `$predict_type` was written to the model even when the `AutoTuner` was not trained.
+
 # mlr3tuning 1.3.0
 
 * feat: Save `ArchiveAsyncTuning` to a `data.table` with `ArchiveAsyncTuningFrozen`.
diff --git a/R/AutoTuner.R b/R/AutoTuner.R
index 9164476b..0981abe9 100644
--- a/R/AutoTuner.R
+++ b/R/AutoTuner.R
@@ -314,6 +314,8 @@ AutoTuner = R6Class("AutoTuner",
     #' @field predict_type (`character(1)`)\cr
     #' Stores the currently active predict type, e.g. `"response"`.
     #' Must be an element of `$predict_types`.
+    #' A few learners already use the predict type during training.
+    #' So there is no guarantee that changing the predict type after tuning and training will have any effect or does not lead to errors.
     predict_type = function(rhs) {
       if (missing(rhs)) {
         return(private$.predict_type)
@@ -322,10 +324,12 @@ AutoTuner = R6Class("AutoTuner",
         stopf("Learner '%s' does not support predict type '%s'", self$id, rhs)
       }
 
-      # Catches 'Error: Field/Binding is read-only' bug
-      tryCatch({
+      self$instance_args$learner$predict_type = rhs
+
+
+      if (!is.null(self$model)) {
         self$model$learner$predict_type = rhs
-      }, error = function(cond){})
+      }
 
       private$.predict_type = rhs
     },
diff --git a/man/AutoTuner.Rd b/man/AutoTuner.Rd
index afb4408b..2624968c 100644
--- a/man/AutoTuner.Rd
+++ b/man/AutoTuner.Rd
@@ -168,7 +168,9 @@ Short-cut to \code{result} from  tuning instance.}
 
 \item{\code{predict_type}}{(\code{character(1)})\cr
 Stores the currently active predict type, e.g. \code{"response"}.
-Must be an element of \verb{$predict_types}.}
+Must be an element of \verb{$predict_types}.
+A few learners already use the predict type during training.
+So there is no guarantee that changing the predict type after tuning and training will have any effect or does not lead to errors.}
 
 \item{\code{hash}}{(\code{character(1)})\cr
 Hash (unique identifier) for this object.}
diff --git a/tests/testthat/test_AutoTuner.R b/tests/testthat/test_AutoTuner.R
index 64e9b9ac..66c85987 100644
--- a/tests/testthat/test_AutoTuner.R
+++ b/tests/testthat/test_AutoTuner.R
@@ -231,22 +231,44 @@ test_that("store_tuning_instance, store_benchmark_result and store_models flags
 })
 
 test_that("predict_type works", {
-  te = trm("evals", n_evals = 4)
-  task = tsk("iris")
-  ps = TEST_MAKE_PS1(n_dim = 1)
-  ms = msr("classif.ce")
-  tuner = tnr("grid_search", resolution = 3)
+  task = tsk("pima")
 
-  at = AutoTuner$new(lrn("classif.rpart"), rsmp("holdout"), ms, te,
-    tuner = tuner, ps)
+  # response predict type
+   at = auto_tuner(
+    tuner = tnr("random_search"),
+    learner = lrn("classif.rpart"),
+    resampling = rsmp("holdout"),
+    measure = msr("classif.ce"),
+    terminator = trm("evals", n_evals = 4))
+
+  expect_equal(at$predict_type, "response")
 
   at$train(task)
   expect_equal(at$predict_type, "response")
   expect_equal(at$model$learner$predict_type, "response")
 
+  # change predict type after training
   at$predict_type = "prob"
   expect_equal(at$predict_type, "prob")
   expect_equal(at$model$learner$predict_type, "prob")
+
+  # prob predict type
+  at = auto_tuner(
+    tuner = tnr("random_search"),
+    learner = lrn("classif.rpart", predict_type = "prob"),
+    resampling = rsmp("holdout"),
+    measure = msr("classif.ce"),
+    terminator = trm("evals", n_evals = 4))
+
+  expect_equal(at$predict_type, "prob")
+
+  at$train(task)
+
+  expect_equal(at$predict_type, "prob")
+  expect_equal(at$model$learner$predict_type, "prob")
+
+  pred = at$predict(task)
+  expect_numeric(pred$score(msr("classif.auc")))
 })
 
 test_that("search space from TuneToken works", {