In an environment increasingly characterised by an avalanche of data and the rapid advancement of generative artificial intelligence (AI), metadata – also known as information about the data – play a critical role. This paper sheds light on their value and importance for AI. First, it explores the relationship between generative AI and metadata, depicting a cycle in which AI acts as both a user and producer of metadata. In this virtuous cycle, AI enhances the quality and availability of metadata, which in turn improves AI by making data more understandable. This dynamic is exemplified through two case studies from the Bank for International Settlements. However, this “virtuous” cycle can also become “vicious” where the dissemination of inaccurate metadata leads to flawed AI, data misuse and, ultimately, misinformation. This underscores the importance of producing high-quality, standardised and machine-readable data, and of providing support to official statisticians in this effort, in spite of increasingly strained resources.
United Nations Economic Commission for Europe (UNECE). Getting the facts right: a guide to presenting metadata. ECE/CES/29. 2013.
2.
DippoCSundgrenB. The role of metadata in statistics. US Bureau of Labor Statistics. Statistical Survey Paper. 2000.
3.
KrizmanITissotB. Data governance: issues for the national statistical system. IFC Bulletin. 2021; 54: 1–27.
4.
FroeschlK. Metadata management in official statistics – an IT-based methodology approach. Austrian J Stat. 1999; 28: 49–79.
5.
Irving Fisher Committee on Central Bank Statistics (IFC). Granular data: new horizons and challenges. IFC Bulletin. 2024; 61.
6.
United Nations Economic Commission for Europe (UNECE). In-depth review of linking data across domains and sources. In: Conference of European Statisticians, 72nd plenary session, Geneva, Switzerland, 20–21 June 2024. ECE/CES/2024/5.
7.
IsraëlJ-MTissotB. Incorporating micro data into macro policy decision-making. J Digital Banking. 2021; 6: 233–250.
8.
LaaboudiCKarlbergMIslamM. Open data dissemination at eurostat: state of the art. Stat J IAOS. 2024; 40: 81–90.
9.
GregoryA. Open and FAIR: trends in scientific publishing and the implications for official statistics. Stat J IAOS.2024; 40: 91–96.
Statistical Data and Metadata Exchange (SDMX). A reference framework for SDMX structural metadata governance. 2023.
12.
WilkinsonMDumontierMAalbersbergIJJ, et al.The FAIR guiding principles for scientific data management and stewardship. Sci Data. 2016; 3: 160018.
13.
United Nations Economic Commission for Europe (UNECE). Data governance framework for statistical interoperability. Report. High-Level Group for the Modernisation of Official Statistics. 2024.
14.
Bank for International Settlements (BIS). Artificial intelligence and the economy: implications for central banks. Annual Economic Report. 2024; 91–127.
15.
United Nations Economic Commission for Europe (UNECE). Data stewardship and the role of national statistical offices in the new data ecosystem. ECE/CES/STAT/2023/4. 2024.
16.
AraujoDDoerrSGambacortaL,et al.Artificial intelligence in central banking. BIS Bulletin. 2024; 84.
17.
United Nations Economic Commission for Europe (UNECE). Large language models for official statistics. White Paper. High-Level Group for the Modernisation of Official Statistics. 2023.
18.
FraislD.The potential of artificial intelligence for the SDGs and official statistics. PARIS21 Working Paper. 2024.
19.
AraujoDProbstSSchimdtR, et al.Harnessing recent breakthroughs in large language models (LLMs). box 2. In: AraujoDBrunoGMarcucciJ, et al. (eds) Data science in central banking: applications and tools. 59. IFC Bulletin. 2023; pp.17–20.
SirelloO.Editing metadata with generative AI. UNECE Expert Meeting on Statistical Data Editing. 2024
27.
Statistical Data and Metadata Exchange (SDMX). SDMX+AI: unlocking the potential of NLP to enhance data access. 2024.
28.
Irving Fisher Committee on Central Bank Statistics (IFC). SDMX adoption and use of open source tools. IFC Report. 2025; 17.
29.
LambeEParkT. The BIS data portal project – delivering the next generation platform for BIS statistics. IFC Bulletin. 2024; 60.
30.
WeiJWangXSchuurmansD, et al.Chain-of-thought prompting elicits reasoning in large language models. In: KoyejoS, et al. (eds) Advances in neural information processing systems. 2022; 35: 24824–24837.
31.
United Nations Economic Commission for Europe (UNECE). Organisational aspects of implementing ML based data editing in statistical production. High-Level Group for the Modernisation of Official Statistics. 2024.
32.
United Nations Economic Commission for Europe (UNECE). Mitigating and monitoring in generative artificial intelligence in official statistics. 2024.
33.
GarrettA. The devil, the detail, and the data. Journal of the Royal Statistical Society Series A. 2024; 187: 857–878.
34.
Organisation for Economic Co-operation and Development (OECD). Embracing the technology frontier. OECD Digital Economy Outlook. 2024.
35.
AllenB.The role of metadata in the second machine age. In: International Conference on Dublin Core and Metadata Applications. Copenhagen. Denmark. 2016.
36.
Irving Fisher Committee on Central Bank Statistics (IFC). Governance and implementation of artificial intelligence in central banks. IFC Report. 2025;18.
37.
US Department of Commerce. Generative artificial intelligence and open data: guidelines and best practices. AI and Open Government Data Assets Working Group. 2025.
38.
Irving Fisher Committee on Central Bank Statistics (IFC). Governance and implementation of AI in central banks. IFC Report. 2025; 18.
