Google’s MCP Server Aims to Reduce AI Hallucinations with Verifiable Data

Google has unveiled its Model Context Protocol (MCP) Server for the Data Commons platform, designed to mitigate AI hallucinations by providing access to verified public datasets. The MCP Server enables developers and AI systems to query census figures, climate data, and UN statistics using natural language, streamlining integration into AI training pipelines.

The Data Commons platform, launched in 2018, consolidates datasets from governments and international organizations. Previously, technical expertise was required to access this data. The MCP Server eliminates that barrier, allowing AI agents to generate insights from structured sources rather than unverified web content—a key step toward reducing generative AI inaccuracies.

A demonstration video shows the server connecting Google Analytics data to large language models like Gemini, enabling marketing plan generation in minutes. This development marks a shift from training AI on unreliable web scrapes to curated, authoritative datasets.