Google AI Overview MCP server

🔎 Google AI Overview API: Get Google's AI answer and its sources as JSON

The most efficient, reliable, and developer-friendly way to use the Google AI Overview API.

Actor page: apify.com/johnvc/Google-AI-Overview-API Input schema: apify.com/johnvc/Google-AI-Overview-API/input-schema

Send any search query and get back Google's AI Overview, the AI-generated answer that appears at the top of Google's results, as structured JSON: the answer text blocks plus the sources it cites (title, link, snippet, and source name). When Google defers generation and returns only a short-lived token, this API resolves the full answer for you automatically, so you always get clean, ready-to-use data.

Video Walkthrough

Quick Start

Prerequisites

• Python 3.11 or higher
• An Apify account and API key (get a free key here)

1. Clone the repository

git clone https://github.com/johnisanerd/Apify-Google-AI-Overview-API.git
   cd Apify-Google-AI-Overview-API

1. Install dependencies with UV

# Install UV if you do not have it:
   curl -LsSf https://astral.sh/uv/install.sh | sh

   # Install project dependencies:
   uv sync

1. Configure your API key

cp .env.example .env
   # Edit .env and add your Apify API key
   # Get your free API key at: https://apify.com?fpr=9n7kx3

1. Run the example

uv run python google-ai-overview-api-example.py

Alternative: set the API key directly

export APIFY_API_TOKEN="your_api_key_here"
uv run python google-ai-overview-api-example.py

Why Use This Google AI Overview API?

Just the answer, structured. You get the AI Overview content and its cited sources as clean JSON, ready to drop into a database, a report, or an LLM pipeline. No HTML parsing.

It handles Google's deferral for you. Google sometimes returns the AI Overview immediately and sometimes defers it behind a short-lived token. This API resolves that follow-up automatically in a single run, so you never have to manage tokens that expire in under a minute.

One query or many. Pass a single query or a queries list. You get one result row per query, which makes batch lookups and GEO/AEO monitoring straightforward.

Pay only for what you fetch. Pricing is per retrieval: one retrieval for a standard inline answer, two only when Google defers and a follow-up is required. The output tells you exactly how many were used.

Country and language targeting. Choose the country (gl) and language (hl), or narrow to a specific location.

MCP-ready. Load it as a tool in Claude Cowork, Claude Code, Claude on the web, and Cursor (see below) and ask for an AI Overview in plain language.

Features

Core Capabilities

• Returns the AI Overview answer as ordered text_blocks (paragraphs, headings, lists)
• Returns the cited references for each answer (title, link, snippet, source)
• Automatic resolution of Google's deferred (token-based) generation
• Single-query and batch (queries) input, one row per query
• Country, language, and location targeting

Data Quality

• Clean, typed JSON output, one row per query, with a stable result_type field
• A clear aioverviewpresent flag and a note when Google shows no overview
• Transparent retrievals_used count so cost is never a surprise

Usage Examples

Basic Example

{
  "query": "what is the Siliconimist podcast?"
}

Advanced Example

{
  "queries": [
    "best practices for prompt engineering",
    "how does vector search work"
  ],
  "gl": "us",
  "hl": "en",
  "location": "Austin, Texas, United States"
}

Shortened. The full README is on GitHub.