diff --git a/tests/test_incremental.py b/tests/test_incremental.py
index 2e2e902e..69b226d8 100644
--- a/tests/test_incremental.py
+++ b/tests/test_incremental.py
@@ -1,6 +1,7 @@
 """Integration tests for incremental graphify extract behavior."""
 from __future__ import annotations
 import json
+import os
 import subprocess
 import sys
 from pathlib import Path
@@ -9,13 +10,25 @@ import pytest
 
 PYTHON = sys.executable
 
+# Backend-selecting env vars. These tests assume no working LLM backend (a docs
+# corpus should fail without one); strip them so a developer who has a real
+# ANTHROPIC_API_KEY / OPENAI_API_KEY / etc. exported does not make a docs extract
+# succeed and break the "no backend" path. CI has none of these set anyway.
+_LLM_ENV_KEYS = (
+    "ANTHROPIC_API_KEY", "OPENAI_API_KEY", "GEMINI_API_KEY", "GOOGLE_API_KEY",
+    "MOONSHOT_API_KEY", "DEEPSEEK_API_KEY", "OLLAMA_BASE_URL",
+    "AWS_PROFILE", "AWS_REGION", "AWS_DEFAULT_REGION", "AWS_ACCESS_KEY_ID",
+)
+
 
 def _run(args: list[str], cwd: Path) -> subprocess.CompletedProcess:
+    env = {k: v for k, v in os.environ.items() if k not in _LLM_ENV_KEYS}
     return subprocess.run(
         [PYTHON, "-m", "graphify"] + args,
         cwd=cwd,
         capture_output=True,
         text=True,
+        env=env,
     )