StatGPT: IMF’s New AI Tool Aims to Protect Accuracy of Global Official Statistics

The IMF and global statistical institutions propose StatGPT, an AI system that retrieves official statistics directly from trusted databases using natural language queries, ensuring accuracy instead of generating numbers like typical chatbots. The initiative aims to make official data easier to find while protecting the reliability and authority of global statistical systems.

As artificial intelligence becomes a common way to search for information, global statistical institutions are asking a critical question: how can AI help people access official statistics without spreading inaccurate numbers? A new study from the International Monetary Fund’s Statistics Department explores this challenge and proposes a solution called StatGPT. The research, written by IMF economists James Tebrake, Bachir Boukherouaa, Jeff Danforth, and Niva Harikrishnan, reflects broader collaboration among leading institutions such as the IMF, the World Bank, the Organisation for Economic Co-operation and Development (OECD), Eurostat, the Bank for International Settlements (BIS), the European Central Bank (ECB), and the United Nations statistical system. Together, these organizations manage vast databases of official economic and social statistics used by governments, researchers, and policymakers worldwide.

Why Official Statistics Matter

Official statistics form the backbone of public policy and economic decision-making. Governments rely on them to track economic growth, inflation, employment, trade, and population trends. These numbers are produced by national statistical offices using internationally recognized methods so that data are reliable and comparable across countries.

Despite this rigorous process, many people still struggle to find the statistics they need. The problem is not that data are unavailable. Instead, the challenge lies in navigating complicated databases and identifying the correct indicators. Researchers often spend hours searching across different websites and data portals before locating the exact figure required for analysis.

The Challenge of Finding the Right Data

Several obstacles make official statistics difficult to access. One of the biggest is data discovery. When users search for information online, they encounter multiple sources, including national statistical offices, international organizations, and private data platforms. Determining which source contains the official number can be confusing.

Language differences also create barriers. People typically search using everyday terms such as “inflation” or “economic growth.” However, statistical databases often use technical names like “Consumer Price Index, All Items” or “GDP, Constant Prices.” Because of this mismatch, users may overlook the correct dataset or select indicators that do not precisely answer their question.

Another complication is how statistics are organized. Databases are usually divided into categories such as national accounts, prices, or labor markets. Yet real-world questions often require information from several of these categories at once. A researcher studying economic trends may need GDP, employment, inflation, and population data simultaneously. Collecting these figures often requires downloading multiple datasets and combining them manually.

The Risk of AI Generating “Plausible but Wrong” Numbers

Generative AI tools like chatbots appear to offer a solution by allowing users to ask questions in natural language and receive instant answers. In theory, this could make statistical data much easier to access. A user could simply ask an AI system for economic growth rates across countries and receive a table within seconds.

However, the IMF research warns that current AI models often produce numbers that look correct but are actually inaccurate. When tested against official economic data, AI-generated responses frequently contained errors, even though the values appeared reasonable. Because these figures seem believable, users may accept them without verifying the source.

This poses a serious risk for policymakers and researchers. Official statistics require exact precision. Even small differences in numbers can affect economic analysis and policy decisions. If AI tools provide inaccurate data, trust in both the technology and the statistical system could be undermined.

StatGPT: Connecting AI to Official Databases

To address this problem, the IMF researchers propose StatGPT, a new approach that combines conversational AI with direct access to official statistical databases. Unlike typical chatbots, StatGPT does not generate numbers on its own. Instead, it interprets a user’s question and converts it into a structured query sent to official data systems.

Once the query is processed, the system retrieves the exact figures published by the statistical authority and presents them to the user in an easy-to-read format. In simple terms, the AI understands the question, but the numbers come directly from the official database.

Users can also refine their requests during the conversation by specifying countries, time periods, or indicators. Before retrieving the data, the system displays the interpreted query so users can confirm that it reflects their request accurately.

Building an AI-Ready Statistical System

Integrating AI into official statistics will require changes in how data are published. Many statistical agencies still provide information through downloadable spreadsheets or web tables designed for human browsing. To support AI tools, data must be accessible through standardized application programming interfaces and supported by detailed metadata explaining each indicator.

International data standards, such as the Statistical Data and Metadata Exchange framework, already provide a foundation for this transformation. With these systems in place, AI tools like StatGPT could connect directly to trusted statistical sources around the world.

As AI becomes the default way people search for information, ensuring the reliability of official statistics is more important than ever. By combining the speed of AI with the authority of official databases, initiatives like StatGPT aim to make trustworthy data easier to find while preserving the integrity of global statistical systems.