diff --git a/TECHDEBTS.md b/TECHDEBTS.md
index 8e17dbf..2284f8e 100644
--- a/TECHDEBTS.md
+++ b/TECHDEBTS.md
@@ -664,7 +664,16 @@ user projects.
 **Acceptance.** CRLF fixture parses; round-trip property tests pass; golden
 serialization fixtures unchanged (if keeping custom serializer).
 
-### TD-17: Language detection via `paasaa` (optional, low priority)
+### TD-17: Language detection via `paasaa` (optional, low priority) ✅ DONE (2026-06-12)
+
+**Status: implemented without adding `paasaa`.** The originally reported
+misclassifications are not reproducible on the current code: the existing
+detector already classifies the relevant umlaut-free German and accent-free
+French fixtures correctly through its language-hint fallback, and new focused
+tests now lock that behavior down directly. Because the acceptance cases are now
+satisfied and the current implementation keeps dependency weight lower, the
+project does not add `paasaa` at this time. Revisit only if broader real-world
+fixtures start failing.
 
 **Context.** `Search.detect_language/1` uses diacritic regexes + tiny word
 lists; German text without umlauts (common in short posts) falls through to
diff --git a/test/bds/search_test.exs b/test/bds/search_test.exs
index 5261e06..87375f5 100644
--- a/test/bds/search_test.exs
+++ b/test/bds/search_test.exs
@@ -553,6 +553,15 @@ defmodule BDS.SearchTest do
     assert Enum.uniq(languages) == languages
   end
 
+  test "detect_language classifies umlaut-free German text as German" do
+    assert BDS.Search.detect_language("Der Fluss fliesst ruhig am Morgen entlang der alten Bruecke") ==
+             "de"
+  end
+
+  test "detect_language classifies accent-free French text as French" do
+    assert BDS.Search.detect_language("Je cours au parc chaque matin avant le travail") == "fr"
+  end
+
   test "search_posts finds translation text in multiple languages after reindex", %{
     project: project
   } do