diff --git a/Utils/Utility.py b/Utils/Utility.py
index d3644d2..74ac60c 100644
--- a/Utils/Utility.py
+++ b/Utils/Utility.py
@@ -18,6 +18,17 @@
 _NONSPACE_PATTERN = re.compile(r"\S+")
 _SLUG_PATTERN = re.compile(r"[^a-z0-9]+")
 _AUTHOR_CASE_PATTERN = re.compile("^\\w| \\w")
+# A handful of WordPress accounts were registered with the byline text as the
+# login ("By Nayab Iqbal"), so the slug built from that login carries a "by-"
+# the display name never had. _AUTHOR_CLEAN_PATTERN already strips this from
+# names; it cannot be reused on a slug, because it also strips digits and would
+# rewrite the numeric suffix dedupe_slug adds to break collisions.
+#
+# Anchored and dash-terminated on purpose: it must not touch a surname that
+# begins with those letters, and canonicalize_slug has already collapsed the
+# separator, so "By Nayab Iqbal" and "by_nayab_iqbal" both arrive as
+# "by-nayab-iqbal" while "Byrne" arrives as "byrne" and is left alone.
+_AUTHOR_SLUG_BY_PREFIX = re.compile(r"^by-")
 
 class Utility:
   def _obj_data(item):
@@ -51,6 +62,7 @@ def canonicalizeAuthorLogins(authors):
       data = Utility._obj_data(author)
       author_id = data.get("id")
       base = Utility.canonicalize_slug(data.get("login")) or Utility.canonicalize_slug(data.get("display_name"))
+      base = _AUTHOR_SLUG_BY_PREFIX.sub("", base)
       data["login"] = Utility.dedupe_slug(used, base, "author", author_id)
 
   def canonicalizeArticleSlugs(articles):
diff --git a/tests/test_author_login_slugs.py b/tests/test_author_login_slugs.py
new file mode 100644
index 0000000..8dbde09
--- /dev/null
+++ b/tests/test_author_login_slugs.py
@@ -0,0 +1,72 @@
+"""Tests for the author slug built from the WordPress login.
+
+Run from the repo root:
+
+    .venv/bin/python -m unittest tests.test_author_login_slugs
+"""
+import os
+import sys
+import unittest
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+from Utils.Utility import Utility
+
+
+def author(author_id, login, display_name=None):
+    return {"id": author_id, "login": login, "display_name": display_name}
+
+
+def logins(authors):
+    Utility.canonicalizeAuthorLogins(authors)
+    return [a["login"] for a in authors]
+
+
+class AuthorLoginSlugs(unittest.TestCase):
+    def test_by_prefix_from_the_login_is_dropped(self):
+        # The regression: a few accounts were registered with the byline text
+        # as the login, so the slug carried a "by-" the display name never had
+        # and /author/nayab-iqbal 404'd while /author/by-nayab-iqbal served.
+        self.assertEqual(
+            logins([author(1, "By Nayab Iqbal", "Nayab Iqbal")]),
+            ["nayab-iqbal"],
+        )
+
+    def test_separator_spelling_does_not_matter(self):
+        # canonicalize_slug collapses the separator before the prefix is
+        # stripped, so every spelling of the login arrives the same way.
+        self.assertEqual(logins([author(1, "by-lena-tran")]), ["lena-tran"])
+        self.assertEqual(logins([author(2, "by_jack_davis")]), ["jack-davis"])
+        self.assertEqual(logins([author(3, "BY ANUM HASSAN")]), ["anum-hassan"])
+
+    def test_a_surname_beginning_with_by_is_left_alone(self):
+        # The prefix is dash-terminated precisely so this keeps working.
+        self.assertEqual(logins([author(1, "Byrne")]), ["byrne"])
+        self.assertEqual(logins([author(2, "byers")]), ["byers"])
+
+    def test_falls_back_to_the_display_name_and_still_strips(self):
+        self.assertEqual(
+            logins([author(1, None, "By Jack Davis")]),
+            ["jack-davis"],
+        )
+
+    def test_stripping_can_collide_and_still_dedupes(self):
+        # "By Jack Davis" and "Jack Davis" both reduce to jack-davis, so the
+        # second must still be given a distinct slug rather than silently
+        # overwriting the first.
+        self.assertEqual(
+            logins([author(10, "jack-davis"), author(11, "by-jack-davis")]),
+            ["jack-davis", "jack-davis-11"],
+        )
+
+    def test_stripping_never_empties_a_slug(self):
+        # "by-" canonicalizes to "by" before the pattern is applied, and the
+        # pattern needs the trailing dash, so nothing is stripped and the slug
+        # stays non-empty. Worth pinning: a rule that could reduce a login to ""
+        # would send every such author through the same author-<id> fallback.
+        self.assertEqual(logins([author(7, "by-")]), ["by"])
+        self.assertEqual(logins([author(8, "By")]), ["by"])
+
+
+if __name__ == "__main__":
+    unittest.main()