Talk to Your Google Ads

Saturday, 10 May 2025 | Automation AI & ML N8N LLM Google Ads
Summary

Query your Google Ads performance and monitor spend anytime through a simple chat interface, completely eliminating complex dashboards.

Background

Managing digital marketing campaigns typically requires navigating notoriously dense, complex interfaces. For busy business owners and marketers, pulling basic metrics from Google Ads—such as daily spend pacing, current Return on Ad Spend (ROAS), or identifying underperforming keywords—means clicking through multiple layers of nested menus or building fragile custom dashboards. This structural friction creates operational bottlenecks; studies on data accessibility show that professionals waste countless hours generating routine performance reports, leading to delayed decisions and missed optimization windows.

To solve this, modern data ecosystems are rapidly shifting toward Conversational Business Intelligence (Conversational BI). According to market insights on AI-driven data analytics trends, the integration of natural language processing with live enterprise databases allows teams to bypass traditional UI barriers completely. Instead of logging into a heavy dashboard to build a custom query, users can simply ask an AI assistant questions in plain language. This project emerges to dismantle reporting friction by wrapping the Google Ads reporting API into a responsive chat interface, making campaign diagnostics as instantaneous as sending a text message.

Goals

By implementing this project, we aim to achieve the following:

Instantaneous Reporting

Fetch high-level or granular performance metrics (e.g., impressions, clicks, conversions, ROAS) on-demand through real-time text prompts.

Effortless Financial Oversight

Monitor daily and monthly ad spend pacing continuously to instantly detect and prevent overspending or budget bleeding.

Rapid Actionable Insights

Quickly identify what is working (winning creatives, high-performing asset groups) and what isn't, without digging through spreadsheets.

Democratize Campaign Data

Remove technical barriers to entry, enabling any non-technical teammate or stakeholders to safely check ad health on the fly.

Approach

To build an on-demand reporting layer for Google Ads, we evaluated three primary options: standard Looker Studio automated dashboards, third-party scheduled email reports, and an interactive AI agent. Static dashboards and scheduled emails are passive and rigid; they cannot answer unplanned, specific questions in the moment (e.g., "How much did we spend on brand search campaigns this specific morning?").

Having an AI agent is the best approach to handle the dynamic of the questions arise. Several identified option to have a chat with AI Agent including: Deploying an Google Ads MCP as a terminal, having a conversation using Claude native connectors, and deploy our own manual VPS and workflow. The first option may not look practical for non IT users while the second option may incur higher cost and lack of flexibility since it's attached to the Claude Code. Deploying our own MCP and setting the workflow may seems complicated at the beginning, but it allows multi channel integration, freedom of AI model selection for better cost-performance, & omniconversation with any team members within the company

The best option is an LLM-driven conversational agent orchestrated by n8n. By utilizing an LLM featuring advanced tool-calling/function-calling capabilities, the AI acts as a smart interpreter. When a user sends a chat message like "Which campaign had the lowest CPA last week?", n8n captures the intent, converts the natural language request into the correct Google Ads Query Language (GAQL) request, fetches the live data from the Google Ads API, and hands it back to the LLM to write a concise, human-friendly summary. This creates a flexible, active reporting cycle completely customized to the user's sudden inputs.

Scalability

This website is hosted on a VPS with a minimum requirement since least memory and storage are needed to run the website platform. For higher interaction intensity, you want to scale up the RAM and vCPU to meet your daily needs. Currently, this MCP settings only applies one account per server. For multiple account interaction, you may want to modify the official MCP Google Ads on Github so that an account can be pass via conversation. In that case, you also require to use db as memory to store the conversation.

For building out the tracking blueprint, it is highly recommended to reference official GA4 documentation, particularly regarding recommended standard events and custom dimensions. For higher scale data collection and advanced vertical tracking, GA4 provides preset standard events and flexible parameters to measure various typical user journeys out of the box. In those high-scale integration cases, global dataLayer structure requirements may expand due to additional custom parameter needs.

Implementation Workflow

1. Prepare Google Ads Information

Prepare Google Ads Account ID & Google Ads Developer Token.

Ensure the Google Ads account (not the manager) has been included in the MCC account.

Google Dev Token can only be generated by Google Ads MCC (Manager Account). In addition, the developer token at least need to be Basic level.

2. Create an automation flow on n8n

Define the input channel (Slack, Google Chat, Teams, etc) and connect it to an AI Agent.

3. Define AI Agent Instructions

Define an instruction using system prompt for AI Agent for a better context.

4. Configure VPS & Environments

Upload the Google Ads MCP to your VPS.

Upload the Google Ads & Google Cloud Project information within .env.

Download the Official Google Ads MCP to your VPS.

5. Connect AI Agent with MCP Client

Since it supports configuration via FastMCP, we can directly use the HTTP Stream/SSE by connecting it into MCP Client Tool in Tools sub node to your VPS.

6. Test & Publish the workflow

Since it uses OAuth2 access, it will require to login first for authentication. After that, we can test by having an interaction with the bot.

Publish the workflow to the production.

Interested in a similar automation project?

Let's collaborate to map out your goals, design system, or integrate custom intelligent components tailored to your needs.

Contact Me

Requirements

  • Automation Hub (n8n) An active n8n instance to serve as the backend coordinator, handling webhook triggers from your messaging channel and structuring API calls.
  • LLM Account Access to an LLM provider (e.g., OpenAI, Anthropic, or Google Gemini) supporting function-calling to accurately map chat prompts into specific data commands.
  • Google Cloud Project An active Google Cloud Project, including the OAuth2 secret key for authentication.
  • Google Ads Developer Token Valid OAuth2 credentials and a Google Ads Developer Token to authorize secure, read-only requests to your advertising accounts.
  • Messaging Interface Endpoint A connected chat endpoint (such as a private Telegram bot, Discord bot, or Slack App) to serve as your communication interface.

References

  • Conversational BI and Enterprise AI Shifts:
    eWeek Insights, Top Conversational AI Trends. Analysis outlining how modern enterprises are adopting natural language interfaces to democratize complex data collection and eliminate dashboard fatigue.
  • The Evolution of Analytics Interfaces:
    VentureBeat AI Focus, How Generative AI is Rewriting the Rules of Business Dashboards. Technical overview demonstrating how text-to-API architecture accelerates operational speed across marketing teams.
  • Google Developer Token:
    Unique Key used to access Google Ads API. Explaining why it’s important, how to generate and raise the access level.

Related Posts