diff --git a/atlasbot/engine/answerer.py b/atlasbot/engine/answerer.py
index fbcfe16..91ed684 100644
--- a/atlasbot/engine/answerer.py
+++ b/atlasbot/engine/answerer.py
@@ -985,6 +985,7 @@ def _metric_candidate_lines(lines: list[str], keywords: list[str] | None, limit:
     if not lines:
         return []
     lowered = [kw.lower() for kw in (keywords or []) if kw]
+    prefer_node = any("node" in kw for kw in lowered) or "hottest" in lowered
     metric_tokens = {
         "cpu",
         "ram",
@@ -1004,6 +1005,8 @@ def _metric_candidate_lines(lines: list[str], keywords: list[str] | None, limit:
     candidates: list[str] = []
     for line in lines:
         line_lower = line.lower()
+        if prefer_node and "pod_" in line_lower:
+            continue
         if "hottest:" in line_lower:
             candidates.append(line)
             continue
diff --git a/atlasbot/llm/prompts.py b/atlasbot/llm/prompts.py
index 37a6834..f5a2fb8 100644
--- a/atlasbot/llm/prompts.py
+++ b/atlasbot/llm/prompts.py
@@ -215,7 +215,9 @@ FACT_SELECT_SYSTEM = (
 
 FACT_SELECT_PROMPT = (
     "Pick up to {max_lines} lines from Candidates that best answer the question. "
-    "If the question asks for highest/hottest and Candidates include a line starting with 'hottest:', prefer that line. "
+    "If the question asks for highest/hottest and Candidates include a line starting with 'hottest:', you must include that line. "
+    "If the question mentions nodes and a 'hottest:' line exists, prefer node-level facts over pod-level lines. "
+    "Avoid pod_* lines unless the question explicitly mentions pods. "
     "Return JSON with field: lines (list of strings). If none apply, return {{\"lines\": []}}."
 )