diff --git a/csv_metadata_quality/app.py b/csv_metadata_quality/app.py
index d025be9..36d8aa7 100644
--- a/csv_metadata_quality/app.py
+++ b/csv_metadata_quality/app.py
@@ -21,6 +21,12 @@ def parse_args(argv):
         "-a",
         help="Comma-separated list of fields to validate against AGROVOC, for example: dcterms.subject,cg.coverage.country",
     )
+    parser.add_argument(
+        "--drop-invalid-agrovoc",
+        "-d",
+        help="After validating metadata values against AGROVOC, drop invalid values.",
+        action="store_true",
+    )
     parser.add_argument(
         "--experimental-checks",
         "-e",
@@ -123,12 +129,14 @@ def run(argv):
         # Fix: duplicate metadata values
         df[column] = df[column].apply(fix.duplicates, field_name=column)
 
-        # Check: invalid AGROVOC subject
+        # Check: invalid AGROVOC subject and optionally drop them
         if args.agrovoc_fields:
             # Identify fields the user wants to validate against AGROVOC
             for field in args.agrovoc_fields.split(","):
                 if column == field:
-                    df[column].apply(check.agrovoc, field_name=column)
+                    df[column] = df[column].apply(
+                        check.agrovoc, field_name=column, drop=args.drop_invalid_agrovoc
+                    )
 
         # Check: invalid language
         match = re.match(r"^.*?language.*$", column)
diff --git a/csv_metadata_quality/check.py b/csv_metadata_quality/check.py
index 6a2f89f..7de9374 100755
--- a/csv_metadata_quality/check.py
+++ b/csv_metadata_quality/check.py
@@ -188,7 +188,7 @@ def language(field):
     return
 
 
-def agrovoc(field, field_name):
+def agrovoc(field, field_name, drop):
     """Check subject terms against AGROVOC REST API.
 
     Function constructor expects the field as well as the field name because
@@ -219,6 +219,9 @@ def agrovoc(field, field_name):
     # prune old cache entries
     requests_cache.remove_expired_responses()
 
+    # Initialize an empty list to hold the validated AGROVOC values
+    values = list()
+
     # Try to split multi-value field on "||" separator
     for value in field.split("||"):
         request_url = "http://agrovoc.uniroma2.it/agrovoc/rest/v1/agrovoc/search"
@@ -231,9 +234,25 @@ def agrovoc(field, field_name):
 
             # check if there are any results
             if len(data["results"]) == 0:
-                print(f"{Fore.RED}Invalid AGROVOC ({field_name}): {Fore.RESET}{value}")
+                if drop:
+                    print(
+                        f"{Fore.GREEN}Dropping invalid AGROVOC ({field_name}): {Fore.RESET}{value}"
+                    )
+                else:
+                    print(
+                        f"{Fore.RED}Invalid AGROVOC ({field_name}): {Fore.RESET}{value}"
+                    )
 
-    return
+                    # value is invalid AGROVOC, but we are not dropping
+                    values.append(value)
+            else:
+                # value is valid AGROVOC so save it
+                values.append(value)
+
+    # Create a new field consisting of all values joined with "||"
+    new_field = "||".join(values)
+
+    return new_field
 
 
 def filename_extension(field):