Skip to main content
User Question: “My traffic has dropped 20% this week. Why?” This workflow shows how an agent uses the MCP to act as a detective and attribute a drop to its root cause.

​
Step 1: Confirm the Magnitude

The agent starts by verifying the drop and finding the exact start date using analytics_time_series. Agent Logic:
  • Call analytics_time_series with days: 30.
  • Identify the “inflection point” where clicks started to diverge from the rolling average.

​
Step 2: Site-Wide or Specific?

Next, the agent determines if the drop is universal or localized to certain segments. Agent Logic:
  • Use compare_periods to check:
    • By Device: Is it a mobile-only drop? (Likely PageSpeed or mobile-usability issue).
    • By Country: Is it geographic? (Likely a local Google algorithm update).
    • By Search Appearance: Is it a loss of Rich Snippets?

​
Step 3: Page/Query Attribution

The agent finds the specific pages or keywords that “lost” the most clicks. Agent Logic:
  • Use seo_lost_queries to identify top losers.
  • Cross-reference with inspect_url for those pages to check for manual actions or indexing issues.

​
Step 4: Outcome

The agent concludes the investigation with an actionable report:
“The 20% drop was concentrated on your Product pages (mobile only) starting on Tuesday. URL inspection shows the pages are still indexed, but PageSpeed scores dropped from 90 to 45. It appears a recent code update slowed down these pages, leading to a ranking decline.”