diff --git a/semeio/workflows/csv_export2/csv_export2.py b/semeio/workflows/csv_export2/csv_export2.py
index 537aade99..6e2b4fd7f 100644
--- a/semeio/workflows/csv_export2/csv_export2.py
+++ b/semeio/workflows/csv_export2/csv_export2.py
@@ -6,6 +6,70 @@
 from fmu import ensemble
 from res.job_queue import ErtScript
 
+DESCRIPTION = """
+CSV_EXPORT2 will export selected Eclipse summary vectors to a CSV file.
+The vector selection is independent of the ``SUMMARY`` keywords in the
+ert config file.
+
+The CSV file will look like:
+
+======== ==== =========== ==== ======
+ENSEMBLE REAL DATE        FOPR FOPT
+======== ==== =========== ==== ======
+iter-0   0    2020-01-01  800  0
+iter-0   0    2020-02-01  1000 365000
+iter-0   1    2020-01-01  700  0
+iter-0   1    2020-01-01  1100 401500
+======== ==== =========== ==== ======
+
+The time frequency must be chosen. If ``raw``, the original timesteps from
+Eclipse is chosen, and it will be individual pr. realization. If ``daily``,
+``monthly``  or  ``yearly`` is chosen, only data at those dates are
+given for all realization. Rate data (e.g.  FOPR) is valid for the given dates,
+but can not be summed up to cumulative data when time interpolation. Cumulative
+columns (f.ex. FOPT) are time-interpolated linearly. See the `documentation on
+fmu-ensemble
+<https://equinor.github.io/fmu-ensemble/usage.html#rate-handling-in-eclipse-summary-vectors>`_
+for more details on rate handling.
+
+Columns are selected by a list of strings, where wildcards characters ``?``
+(matches exactly one character) and ``*`` (matches zero or more characters) can
+be used to select multiple columns.
+
+Column count more than 1000 gives increased probability for problems downstream,
+depending on which applications are put into use. Column count depends on the
+combination of wildcards used in this workflow and the actual vectors that are
+requested in the Eclipse DATA file. A wildcard like``W*`` can in certain cases
+(e.g. Eclipse simulations with 100+ wells) produce thousands of vectors, and can
+then be replaced by something more explicit like ``WOPT*  WGPT* WWPT*``.
+"""  # noqa
+
+EXAMPLES = """
+Example
+-------
+
+Add a file named e.g. ``ert/bin/workflows/QC_CSVEXPORT2`` with the contents::
+
+  MAKE_DIRECTORY <CASEDIR>/share/summary/
+  EXPORT_RUNPATH * | *
+  CSV_EXPORT2 <RUNPATH_FILE> <CASEDIR>/share/summary/<CASE>.csv monthly F* W* TCPU TIMESTEP
+
+(where ``<CASEDIR>`` typically points to ``/scratch/..``). Adjust all three
+lines to your needs.
+
+``EXPORT_RUNPATH`` in the workflow file is added to ensure all realizations and
+all iterations are included in the RUNPATH file.  If you have rerun only a
+subset of your ensemble, the RUNPATH file will only contain those unless this
+statement is included.
+
+Add to your ERT config to have the workflow automatically executed on successful
+runs::
+
+  LOAD_WORKFLOW ../bin/workflows/QC_CSVEXPORT2
+  HOOK_WORKFLOW QC_CSVEXPORT2 POST_SIMULATION
+
+"""  # noqa
+
 
 def csv_exporter(runpathfile, time_index, outputfile, column_keys=None):
     """Export CSV data (summary and parameters) from an EnsembleSet
@@ -43,10 +107,32 @@ def main(args):
 def csv_export_parser():
     """Setup parser"""
     parser = argparse.ArgumentParser()
-    parser.add_argument("runpathfile", type=str)
-    parser.add_argument("outputfile", type=str)
-    parser.add_argument("time_index", type=str, default="monthly")
-    parser.add_argument("column_keys", nargs="+", default=None)
+    parser.add_argument(
+        "runpathfile",
+        type=str,
+        help=(
+            "Path to ERT RUNPATH-file, "
+            "usually the ERT magic variable <RUNPATH> can be used"
+        ),
+    )
+    parser.add_argument(
+        "outputfile",
+        type=str,
+        help="Path to CSV file to be written. The directory pointed to must exist.",
+    )
+    parser.add_argument(
+        "time_index",
+        type=str,
+        default="monthly",
+        help=(
+            "Time interval specifier for the output. "
+            "This argument is passed on to fmu-ensemble, "
+            "supported specifiers are 'raw', 'daily', 'monthly' and 'yearly'"
+        ),
+    )
+    parser.add_argument(
+        "column_keys", nargs="+", default=None, help="List of summary vector wildcards"
+    )
     return parser
 
 
@@ -54,6 +140,8 @@ def csv_export_parser():
 def legacy_ertscript_workflow(config):
     workflow = config.add_workflow(CsvExport2Job, "CSV_EXPORT2")
     workflow.parser = csv_export_parser
+    workflow.description = DESCRIPTION
+    workflow.examples = EXAMPLES
 
 
 def cli():
diff --git a/tests/jobs/csv_export2/test_integration.py b/tests/jobs/csv_export2/test_integration.py
index f9de8bd99..c16005f09 100644
--- a/tests/jobs/csv_export2/test_integration.py
+++ b/tests/jobs/csv_export2/test_integration.py
@@ -1,4 +1,6 @@
 import pytest
+import rstcheck
+
 from semeio.workflows.csv_export2 import csv_export2
 
 from tests.jobs.csv_export2 import conftest
@@ -104,6 +106,15 @@ def test_no_iterations(ert_statoil_test_data):
         )
 
 
+@pytest.mark.parametrize("input_rst", [csv_export2.DESCRIPTION, csv_export2.EXAMPLES])
+def test_valid_rst(input_rst):
+    """
+    Check that the documentation passed through the plugin system is
+    valid rst
+    """
+    assert not list(rstcheck.check(input_rst))
+
+
 def verifyExportedFile(exported_file_name, result_header, result_iter_rel):
     with open(exported_file_name, "r") as exported_file:
         lines = exported_file.readlines()