diff --git a/docs/src/manual/PyTorch.md b/docs/src/manual/PyTorch.md
index 9eb49af..f56d76b 100644
--- a/docs/src/manual/PyTorch.md
+++ b/docs/src/manual/PyTorch.md
@@ -38,27 +38,37 @@ torch.save(model, "saved_pytorch_model.pt")
 MathOptAI uses [PythonCall.jl](https://github.com/JuliaPy/PythonCall.jl) to call
 from Julia into Python.
 
-See [CondaPkg.jl](https://github.com/JuliaPy/CondaPkg.jl) for more control over
-how to link Julia to an existing Python environment. For example, if you have an
-existing Python installation (with PyTorch installed), and it is available in
-the current Conda environment, set:
+To use [`PytorchModel`](@ref) your code must load the `PythonCall` package:
+```julia
+import PythonCall
+```
+
+PythonCall uses [CondaPkg.jl](https://github.com/JuliaPy/CondaPkg.jl) to manage
+Python dependencies. See [CondaPkg.jl](https://github.com/JuliaPy/CondaPkg.jl)
+for more control over how to link Julia to an existing Python environment. For
+example, if you have an existing Python installation (with PyTorch installed),
+and it is available in the current Conda environment, do:
 
 ```julia
 ENV["JULIA_CONDAPKG_BACKEND"] = "Current"
+import PythonCall
 ```
 
-before importing PythonCall.jl. If the Python installation can be found on the
-path and it is not in a Conda environment, set:
+If the Python installation can be found on the path and it is not in a Conda
+environment, do:
 
 ```julia
 ENV["JULIA_CONDAPKG_BACKEND"] = "Null"
+import PythonCall
 ```
 
 If `python` is not on your path, you may additionally need to set
-`JULIA_PYTHONCALL_EXE`, for example, to:
+`JULIA_PYTHONCALL_EXE`, for example, do:
 
 ```julia
 ENV["JULIA_PYTHONCALL_EXE"] = "python3"
+ENV["JULIA_CONDAPKG_BACKEND"] = "Null"
+import PythonCall
 ```
 
 ## Basic example
@@ -67,7 +77,7 @@ Use [`MathOptAI.add_predictor`](@ref) to embed a PyTorch model into a JuMP
 model:
 
 ```@repl
-using JuMP, MathOptAI
+using JuMP, MathOptAI, PythonCall
 model = Model();
 @variable(model, x[1:1]);
 predictor = MathOptAI.PytorchModel("saved_pytorch_model.pt");
@@ -81,7 +91,7 @@ formulation
 Use the `reduced_space = true` keyword to formulate a reduced-space model:
 
 ```@repl
-using JuMP, MathOptAI
+using JuMP, MathOptAI, PythonCall
 model = Model();
 @variable(model, x[1:1]);
 predictor = MathOptAI.PytorchModel("saved_pytorch_model.pt");
@@ -96,7 +106,7 @@ formulation
 Use the `gray_box = true` keyword to embed the network as a nonlinear operator:
 
 ```@repl
-using JuMP, MathOptAI
+using JuMP, MathOptAI, PythonCall
 model = Model();
 @variable(model, x[1:1]);
 predictor = MathOptAI.PytorchModel("saved_pytorch_model.pt");
@@ -112,7 +122,7 @@ Pass a dictionary to the `config` keyword that maps the `Symbol` name of each
 PyTorch layer to a MathOptAI predictor:
 
 ```@repl
-using JuMP, MathOptAI
+using JuMP, MathOptAI, PythonCall
 model = Model();
 @variable(model, x[1:1]);
 predictor = MathOptAI.PytorchModel("saved_pytorch_model.pt");
diff --git a/docs/src/tutorials/pytorch.jl b/docs/src/tutorials/pytorch.jl
index 2f1d149..e43cab1 100644
--- a/docs/src/tutorials/pytorch.jl
+++ b/docs/src/tutorials/pytorch.jl
@@ -11,30 +11,40 @@
 
 # ## Python integration
 
-# MathOptAI uses [PythonCall.jl](https://github.com/JuliaPy/PythonCall.jl)
-# to call from Julia into Python.
+# MathOptAI uses [PythonCall.jl](https://github.com/JuliaPy/PythonCall.jl) to call
+# from Julia into Python.
+
+# To use [`PytorchModel`](@ref) your code must load the `PythonCall` package:
+# ```julia
+# import PythonCall
+# ```
 #
-# See [CondaPkg.jl](https://github.com/JuliaPy/CondaPkg.jl) for more control
-# over how to link Julia to an existing Python environment. For example, if you
-# have an existing Python installation (with PyTorch installed), and it is
-# available in the current Conda environment, set:
+# PythonCall uses [CondaPkg.jl](https://github.com/JuliaPy/CondaPkg.jl) to manage
+# Python dependencies. See [CondaPkg.jl](https://github.com/JuliaPy/CondaPkg.jl)
+# for more control over how to link Julia to an existing Python environment. For
+# example, if you have an existing Python installation (with PyTorch installed),
+# and it is available in the current Conda environment, do:
 #
 # ```julia
 # ENV["JULIA_CONDAPKG_BACKEND"] = "Current"
+# import PythonCall
 # ```
 #
-# before importing PythonCall.jl. If the Python installation can be found on
-# the path and it is not in a Conda environment, set:
+# If the Python installation can be found on the path and it is not in a Conda
+# environment, do:
 #
 # ```julia
 # ENV["JULIA_CONDAPKG_BACKEND"] = "Null"
+# import PythonCall
 # ```
 #
 # If `python` is not on your path, you may additionally need to set
-# `JULIA_PYTHONCALL_EXE`, for example, to:
+# `JULIA_PYTHONCALL_EXE`, for example, do:
 #
 # ```julia
 # ENV["JULIA_PYTHONCALL_EXE"] = "python3"
+# ENV["JULIA_CONDAPKG_BACKEND"] = "Null"
+# import PythonCall
 # ```
 
 # ## Required packages
@@ -46,6 +56,7 @@ using Test
 import Ipopt
 import MathOptAI
 import Plots
+import PythonCall
 
 # ## Training a model
 
diff --git a/src/predictors/PytorchModel.jl b/src/predictors/PytorchModel.jl
index ea94b10..d33470d 100644
--- a/src/predictors/PytorchModel.jl
+++ b/src/predictors/PytorchModel.jl
@@ -12,10 +12,18 @@ A wrapper struct for loading a PyTorch model.
 The only supported file extension is `.pt`, where the `.pt` file has been
 created using `torch.save(model, filename)`.
 
+!!! warning
+    To use [`PytorchModel`](@ref), your code must load the `PythonCall` package:
+    ```julia
+    import PythonCall
+    ```
+
 ## Example
 
 ```jldoctest
-julia> using PythonCall, MathOptAI
+julia> using MathOptAI
+
+julia> using PythonCall  #  This line is important!
 
 julia> predictor = PytorchModel("model.pt");
 ```