diff --git a/CHANGELOG.md b/CHANGELOG.md
index 6bf35b087..4c1a571e5 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -17,6 +17,7 @@
 
 ## Breaking Changes
 
+- [#938](https://github.com/pybop-team/PyBOP/pull/938) - Make SALib an optional dependency and remove `sensitivity_analysis` in favour of using SALib directly.
 - [#942](https://github.com/pybop-team/PyBOP/pull/942) - Adds `evaluate_batch` to the costs and ensures that an `Evaluation` is returned.
 
 # [v26.3](https://github.com/pybop-team/PyBOP/tree/v26.3) - 2026-03-05
diff --git a/docs/installation.rst b/docs/installation.rst
index 50138d81e..990db5047 100644
--- a/docs/installation.rst
+++ b/docs/installation.rst
@@ -43,25 +43,43 @@ For those who prefer to install PyBOP from a local clone of the repository or wi
 In editable mode, changes you make to the source code will immediately affect the PyBOP installation without the need for reinstallation.
 
 Optional Dependencies
------------------
-``plotly`` - For plotting, PyBOP uses plotly. It can be installed with:
+---------------------
+``plotly`` - For plotting, PyBOP uses `plotly <https://plotly.com/python/>`_. It can be installed with:
 
 .. code-block:: console
 
     pip install pybop[plot]
 
-``scikit-fem`` - This is a dependency for the multi-dimensional pybamm models, and can be installed using:
+``salib`` - To compute sensitivities, PyBOP can be paired with the `Sensitivity Analysis Library (SALib) <https://salib.readthedocs.io/en/latest/>`_:
+
+.. code-block:: console
+
+    pip install pybop[salib]
+
+``scikit-fem`` - This is a dependency for the multi-dimensional PyBaMM models, and can be installed using:
 
 .. code-block:: console
 
     pip install pybop[scifem]
 
-``bpx`` - To use the Faraday Institution's Battery Parameter eXchange (BPX) package install the optional requirement:
+``bpx`` - To use the Faraday Institution's `Battery Parameter eXchange (BPX) package <https://bpxstandard.com/>`_:
 
 .. code-block:: console
 
     pip install pybop[bpx]
 
+``ep-bolfi`` - To use Expectation Propagation with Bayesian Optimization for Likelihood-Free Inference (`EP-BOLFI <https://github.com/YannickNoelStephanKuhn/EP-BOLFI>`_):
+
+.. code-block:: console
+
+    pip install pybop[ep-bolfi]
+
+``pyprobe`` - To import data from battery cyclers, use `Python Processing for Battery Experiments (PyProBE) <https://pyprobe.readthedocs.io/en/latest/>`_:
+
+.. code-block:: console
+
+    pip install pybop[pyprobe]
+
 To install all the optional dependencies, the command ``pip install pybop[all]`` is available. For more information on the optional packages, users are directed towards the `pyproject.toml <https://github.com/pybop-team/PyBOP/blob/develop/pyproject.toml>`_.
 
 Verifying Installation
diff --git a/pybop/analysis/sensitivity_analysis.py b/pybop/analysis/sensitivity_analysis.py
deleted file mode 100644
index 49370bd20..000000000
--- a/pybop/analysis/sensitivity_analysis.py
+++ /dev/null
@@ -1,49 +0,0 @@
-from typing import TYPE_CHECKING
-
-from SALib.analyze import sobol
-from SALib.sample.sobol import sample
-
-if TYPE_CHECKING:
-    from pybop.problems.problem import Problem
-
-
-def sensitivity_analysis(
-    problem: "Problem", n_samples: int = 256, calc_second_order: bool = False
-) -> dict:
-    """
-    Computes the parameter sensitivities on the combined cost function using
-    SOBOL analysis from the SALib module [1].
-
-    Parameters
-    ----------
-    problem : pybop.Problem
-        The optimisation problem.
-    n_samples : int, optional
-        Number of samples for SOBOL sensitivity analysis,
-        performs best as order of 2, i.e. 128, 256, etc.
-    calc_second_order : bool, optional
-        Whether to calculate second-order sensitivities.
-
-    References
-    ----------
-    .. [1] Iwanaga, T., Usher, W., & Herman, J. (2022). Toward SALib 2.0:
-            Advancing the accessibility and interpretability of global sensitivity
-            analyses. Socio-Environmental Systems Modelling, 4, 18155.
-            doi:10.18174/sesmo.18155
-
-    Returns
-    -------
-    Sensitivities : dict
-    """
-
-    salib_dict = {
-        "names": problem.parameters.names,
-        "bounds": problem.parameters.get_bounds_array(),
-        "num_vars": len(problem.parameters),
-    }
-
-    # Create samples, compute cost
-    param_values = sample(salib_dict, n_samples)
-    costs = problem.evaluate(param_values).values
-
-    return sobol.analyze(salib_dict, costs, calc_second_order=calc_second_order)
diff --git a/pybop/problems/problem.py b/pybop/problems/problem.py
index 2165fd79d..8cb7f6274 100644
--- a/pybop/problems/problem.py
+++ b/pybop/problems/problem.py
@@ -1,6 +1,5 @@
 import numpy as np
 
-from pybop.analysis.sensitivity_analysis import sensitivity_analysis
 from pybop.costs.base_cost import BaseCost
 from pybop.costs.evaluation import Evaluation
 from pybop.parameters.parameter import Inputs, Parameters
@@ -227,25 +226,6 @@ def get_finite_initial_cost(self):
             raise ValueError("The initial parameter values return an infinite cost.")
         return cost0
 
-    def sensitivity_analysis(
-        self, n_samples: int = 256, calc_second_order: bool = False
-    ) -> dict:
-        """
-        Computes the parameter sensitivities on the combined cost function using
-        SOBOL analysis. See pybop.analysis.sensitivity_analysis for more details.
-
-        Parameters
-        ----------
-        n_samples : int, optional
-            Number of samples for SOBOL sensitivity analysis, performs best as a
-            power of 2, i.e. 128, 256, etc.
-        calc_second_order : bool, optional
-            Whether to calculate second-order sensitivities.
-        """
-        return sensitivity_analysis(
-            problem=self, n_samples=n_samples, calc_second_order=calc_second_order
-        )
-
     @property
     def cost(self):
         return self._cost
diff --git a/pyproject.toml b/pyproject.toml
index 7651442b2..acd729723 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -33,11 +33,11 @@ dependencies = [
     "numpy>=1.26",
     "scipy>=1.12",
     "pints>=0.6.0",
-    "SALib>=1.5",
 ]
 
 [project.optional-dependencies]
 plot = ["plotly>=6"]
+salib = ["SALib>=1.5"]
 scifem = [
   "scikit-fem>=8.1.0" # scikit-fem is a dependency for the multi-dimensional pybamm models
 ]
@@ -51,7 +51,7 @@ ep-bolfi = [
 pyprobe = [
   "PyProBE-Data>=2.5.0;python_version >= '3.11' and python_version < '3.13'"
 ]
-all = ["pybop[plot,scifem,bpx,pyprobe,ep-bolfi]"]
+all = ["pybop[plot,salib,scifem,bpx,pyprobe,ep-bolfi]"]
 
 [dependency-groups]
 docs = [
diff --git a/tests/unit/test_problem.py b/tests/unit/test_problem.py
index 2e15d7a1d..7a9eb5987 100644
--- a/tests/unit/test_problem.py
+++ b/tests/unit/test_problem.py
@@ -226,22 +226,3 @@ def test_problem_construct_with_model_predict(self, parameters, model, dataset):
                 problem_output["Voltage [V]"].data,
                 atol=1e-6,
             )
-
-    def test_parameter_sensitivities(self, simulator, dataset):
-        cost = pybop.MeanAbsoluteError(dataset)
-        problem = pybop.Problem(simulator, cost)
-        n_params = len(problem.parameters)
-        result = problem.sensitivity_analysis(4, calc_second_order=True)
-
-        # Assertions
-        assert isinstance(result, dict)
-        assert "S1" in result
-        assert "ST" in result
-        assert isinstance(result["S1"], np.ndarray)
-        assert isinstance(result["S2"], np.ndarray)
-        assert isinstance(result["ST"], np.ndarray)
-        assert isinstance(result["S1_conf"], np.ndarray)
-        assert isinstance(result["ST_conf"], np.ndarray)
-        assert isinstance(result["S2_conf"], np.ndarray)
-        assert result["S1"].shape == (n_params,)
-        assert result["ST"].shape == (n_params,)