As artificial intelligence and new sources of data continue to transform how statistics are produced, accessed and used, the OECD has issued a comprehensive update to its statistical quality framework that sets out how the Organisation will safeguard the quality, credibility and trustworthiness of its statistics as the data environment evolves.
The new OECD Statistics and Data Quality Assurance Framework covers the full data lifecycle, from identifying and collecting data through to processing, dissemination and reuse, and is the first edition to explicitly incorporate AI into the assessment and management of statistical quality. As AI systems increasingly mediate access to information, ensuring that reliable statistics can be identified and correctly interpreted is becoming more important.
“Reliable, internationally comparable statistics are the foundation of effective policymaking,” OECD Secretary-General Mathias Cormann said. “As new data sources, analytical methods and artificial intelligence transform how statistics are produced and used, this new Framework will help ensure that OECD statistics remain relevant, credible and trusted.”
The changing data landscape extends well beyond AI. Around 25% of the data currently used to produce OECD statistics come from sources beyond traditional official-statistics transmission programmes, including new sources such as geospatial information, mobile positioning data and commercial transactions. Advances in analytical methods are also changing how these data are processed and used.
AI offers significant opportunities to support statistical production, including by classifying information, detecting anomalies, harmonising metadata and improving access to evidence. At the same time, it creates new challenges for statistical quality. AI-generated or manipulated data may enter the production process, while search engines, large language models and other AI systems increasingly mediate how users find, interpret and reuse statistics, sometimes outside their original context.
The Framework establishes that AI should support, rather than replace, professional statistical judgement. It emphasises human oversight, transparency and traceability, as well as the need to understand where data come from and how they have been transformed.
It also highlights the importance of making authoritative statistics “AI-ready”, with standardised, machine-readable metadata, interoperable formats and clear information on sources and methods helping AI systems identify reliable statistics and interpret them more accurately, while preserving their meaning, context and provenance.
Primarily intended to guide quality management across the OECD, the Framework can also serve, alongside the OECD Recommendation on Good Statistical Practice, as a reference for national and international statistical authorities seeking to review, modernise or strengthen their own quality assurance frameworks.
The OECD Statistics and Data Quality Assurance Framework is available here.