diff --git a/ghostcrew/interface/tui.py b/ghostcrew/interface/tui.py
index 467dfb2..91f48ba 100644
--- a/ghostcrew/interface/tui.py
+++ b/ghostcrew/interface/tui.py
@@ -869,10 +869,9 @@ class GhostCrewTUI(App):
 
     def _add_tool_result(self, name: str, result: str) -> None:
         """Display tool execution result"""
-        # Truncate long results
-        if len(result) > 2000:
-            result = result[:2000] + "\n... (truncated)"
-        self._add_message(ToolResultMessage(name, result))
+        # Hide tool output - LLM will synthesize it in its response
+        # This prevents duplication and keeps the chat clean
+        pass
 
     def _show_system_prompt(self) -> None:
         """Display the current system prompt"""
diff --git a/ghostcrew/tools/web_search/__init__.py b/ghostcrew/tools/web_search/__init__.py
index 6eadd10..af7a8d8 100644
--- a/ghostcrew/tools/web_search/__init__.py
+++ b/ghostcrew/tools/web_search/__init__.py
@@ -74,30 +74,28 @@ async def web_search(arguments: dict, runtime: "Runtime") -> str:
 
 
 def _format_results(query: str, data: dict) -> str:
-    """Format Tavily results for LLM consumption."""
+    """
+    Format Tavily results for LLM consumption.
+    
+    Returns a lean format with summary + titles + URLs only.
+    Content snippets are excluded to prevent LLM from regurgitating
+    noisy web scrapes in its response.
+    """
     parts = [f"Search: {query}\n"]
 
-    # Include synthesized answer if available
+    # Include synthesized answer if available (this is the valuable part)
     if answer := data.get("answer"):
         parts.append(f"Summary:\n{answer}\n")
 
-    # Include individual results
+    # Include sources as title + URL only (no content snippets)
     results = data.get("results", [])
     if results:
         parts.append("Sources:")
         for i, result in enumerate(results, 1):
             title = result.get("title", "Untitled")
             url = result.get("url", "")
-            content = result.get("content", "")
-
-            # Truncate content if too long
-            if len(content) > 500:
-                content = content[:500] + "..."
-
-            parts.append(f"\n[{i}] {title}")
-            parts.append(f"    URL: {url}")
-            if content:
-                parts.append(f"    {content}")
+            parts.append(f"  [{i}] {title}")
+            parts.append(f"      {url}")
     else:
         parts.append("No results found")