diff --git a/src/fev/task.py b/src/fev/task.py
index 6eb995c..a1c73d4 100644
--- a/src/fev/task.py
+++ b/src/fev/task.py
@@ -340,6 +340,11 @@ class Task:
         name of 2 parent directories for local or S3-based datasets.
 
         This field is only here for convenience and is not used for any validation when computing the results.
+    dataset_description : str | None, default None
+        Text description of the dataset.
+    column_descriptions : dict[str, str] | None, default None
+        Text description of each column used by the task. If provided, the keys must exactly match the target,
+        dynamic and static columns of the task.
 
     Examples
     --------
@@ -379,6 +384,8 @@ class Task:
     past_dynamic_columns: list[str] = dataclasses.field(default_factory=list)
     static_columns: list[str] = dataclasses.field(default_factory=list)
     task_name: str | None = None
+    dataset_description: str | None = None
+    column_descriptions: dict[str, str] | None = None
 
     def __post_init__(self):
         if self.task_name is None:
@@ -458,6 +465,16 @@ def __post_init__(self):
                 "`generate_univariate_targets_from` cannot be used for multivariate tasks (when `target` is a list)"
             )
 
+        if self.column_descriptions is not None:
+            expected = set(self.target_columns + self.dynamic_columns + self.static_columns)
+            missing = sorted(expected - set(self.column_descriptions))
+            unexpected = sorted(set(self.column_descriptions) - expected)
+            if missing or unexpected:
+                raise ValueError(
+                    "`column_descriptions` must have exactly one entry per target, dynamic and static column of the "
+                    f"task. Missing: {missing}. Unexpected: {unexpected}."
+                )
+
         # Attributes computed after the dataset is loaded
         self._full_dataset: datasets.Dataset | None = None
         self._freq: str | None = None
diff --git a/test/test_task.py b/test/test_task.py
index eeef7b8..d446371 100644
--- a/test/test_task.py
+++ b/test/test_task.py
@@ -510,3 +510,48 @@ def test_when_datasets_prefix_is_set_then_s3_dataset_is_loaded_from_mirror(tmp_p
     assert loaded_dataset[0]["id"] == "series_0"
     assert task.dataset_path == dataset_path
     assert task.to_dict()["dataset_path"] == dataset_path
+
+
+def test_when_descriptions_not_provided_then_they_default_to_none():
+    task = fev.Task(dataset_path="my_dataset", horizon=12)
+    assert task.dataset_description is None
+    assert task.column_descriptions is None
+    assert task.to_dict()["dataset_description"] is None
+    assert task.to_dict()["column_descriptions"] is None
+
+
+def test_when_column_descriptions_match_task_columns_then_task_is_created():
+    column_descriptions = {"OT": "oil temperature", "HULL": "load", "LULL": "load", "store": "store id"}
+    task = fev.Task(
+        dataset_path="my_dataset",
+        horizon=12,
+        target="OT",
+        known_dynamic_columns=["HULL"],
+        past_dynamic_columns=["LULL"],
+        static_columns=["store"],
+        dataset_description="Transformer data.",
+        column_descriptions=column_descriptions,
+    )
+    assert task.dataset_description == "Transformer data."
+    assert task.column_descriptions == column_descriptions
+    assert fev.Task(**task.to_dict()) == task
+
+
+@pytest.mark.parametrize(
+    "column_descriptions",
+    [
+        {"OT": "oil temperature"},
+        {"OT": "oil temperature", "HULL": "load", "LULL": "load"},
+        {"target": "oil temperature", "HULL": "load"},
+        {},
+    ],
+)
+def test_when_column_descriptions_do_not_match_task_columns_then_validation_error_is_raised(column_descriptions):
+    with pytest.raises(pydantic.ValidationError, match="column_descriptions"):
+        fev.Task(
+            dataset_path="my_dataset",
+            horizon=12,
+            target="OT",
+            known_dynamic_columns=["HULL"],
+            column_descriptions=column_descriptions,
+        )