diff --git a/docker-compose.yml b/docker-compose.yml
index 4cc7605..9907dbf 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -71,7 +71,7 @@ services:
       # serialised under load and pushed TTS to 7-8s; GPU does ~0.3s/sentence.
       MELO_DEVICE: ${MELO_DEVICE:-cuda}
       # Optional single-language lock for replies (empty = user's own language).
-      OUTPUT_LANGUAGE: ${OUTPUT_LANGUAGE:-}
+      OUTPUT_LANGUAGE: ${OUTPUT_LANGUAGE:-ko}
       # Drop the pre-loop planner LLM call to cut voice-reply latency on small
       # hardware (the planner adds a full model round-trip per turn).
       PLANNER_ENABLED: ${PLANNER_ENABLED:-0}
diff --git a/src/jarvis/reply/engine.py b/src/jarvis/reply/engine.py
index 9cc3db3..2ddfe99 100644
--- a/src/jarvis/reply/engine.py
+++ b/src/jarvis/reply/engine.py
@@ -825,6 +825,70 @@ def _build_enrichment_context_hint(cfg, recent_messages: list) -> Optional[str]:
     return "\n\n".join(parts) if parts else None
 
 
+# Site tokens (proper nouns, not language patterns) → controlBrowser search site.
+_SITE_TOKEN_MAP = {
+    "네이버": "naver", "naver": "naver",
+    "구글": "google", "google": "google",
+    "유튜브": "youtube", "유투브": "youtube", "youtube": "youtube",
+    "다음": "daum", "daum": "daum",
+    "빙": "bing", "bing": "bing",
+}
+# Search / open intent words (Korean deployment + English). Kept explicit because
+# this is a DETERMINISTIC fast-path — the small chat model can't be trusted to
+# emit the controlBrowser call reliably, so when the user names a site AND
+# expresses a search/open intent we run it directly, no LLM judgement.
+_SEARCH_INTENT_WORDS = (
+    "검색해줘", "검색해", "검색", "찾아줘", "찾아봐", "찾아", "열어줘", "열어",
+    "들어가줘", "들어가", "띄워줘", "띄워", "보여줘",
+    "search for", "search", "look up", "find", "open", "go to", "navigate",
+)
+
+
+def _maybe_deterministic_site_search(text: str, db, cfg, language) -> Optional[str]:
+    """When broadcasting AND the user names a site AND asks to search/open it,
+    run controlBrowser.search directly so the result actually appears on screen.
+
+    The 3B chat model reliably narrates ("검색하겠습니다") instead of emitting the
+    controlBrowser tool call, so site-specified search is executed
+    deterministically here rather than left to the model. Fail-open: any problem
+    returns None and the normal reply flow continues.
+    """
+    try:
+        from . import turn_state
+        if not getattr(cfg, "stream_browser", True):
+            return None
+        if not turn_state.get_broadcasting():
+            return None
+        low = (text or "").lower()
+        site = tok = None
+        for _t, _key in _SITE_TOKEN_MAP.items():
+            if _t in low:
+                site, tok = _key, _t
+                break
+        if not site or not any(w in low for w in _SEARCH_INTENT_WORDS):
+            return None
+        import re
+        q = re.sub(re.escape(tok) + r"(에서|에다가|에다|에|로|를|을)?", " ", text, flags=re.IGNORECASE)
+        for w in sorted(_SEARCH_INTENT_WORDS, key=len, reverse=True):
+            q = re.sub(re.escape(w), " ", q, flags=re.IGNORECASE)
+        q = re.sub(r"\s+", " ", q).strip(" .,!?。")
+        if not q:
+            q = text
+        from ..tools.registry import run_tool_with_retries
+        res = run_tool_with_retries(
+            db=db, cfg=cfg, tool_name="controlBrowser",
+            tool_args={"action": "search", "site": site, "query": q},
+            system_prompt="", original_prompt="", redacted_text=redact(text),
+            max_retries=1, language=language,
+        )
+        if res and getattr(res, "success", False):
+            debug_log(f"deterministic site search executed: {site} '{q}'", "tools")
+            return res.reply_text or f"{site}에서 '{q}'를 검색해 화면에 띄웠습니다."
+    except Exception as e:  # noqa: BLE001
+        debug_log(f"deterministic site search failed (fail-open): {e}", "tools")
+    return None
+
+
 def run_reply_engine(db: "Database", cfg, tts: Optional[Any],
                     text: str, dialogue_memory: "DialogueMemory",
                     language: Optional[str] = None) -> Optional[str]:
@@ -849,6 +913,13 @@ def run_reply_engine(db: "Database", cfg, tts: Optional[Any],
     # Step 1: Redact sensitive information
     redacted = redact(text)
 
+    # Step 0.5: Deterministic on-screen site search. If the user named a site and
+    # asked to search/open it while broadcasting, do it directly — the small chat
+    # model otherwise just narrates without calling the browser tool.
+    _site_search_reply = _maybe_deterministic_site_search(text, db, cfg, language)
+    if _site_search_reply is not None:
+        return _site_search_reply
+
     # Step 2: Check for recent dialogue context
     recent_messages = []
     is_new_conversation = True