By Niamh Teather
An index only measures what it can see
Every global index depends on the evidence its researchers can see. For the Government AI Readiness Index, that means tracing national strategies, legislation, budgets, institutions and public programmes across nearly 200 countries. Our researchers cross-check what they find, but they cannot recover documents that are poorly indexed, filed under unfamiliar terms or stored in places they do not know to search.
Equivalent policies are not equally visible. A regulatory sandbox announced in English by a ministry with a press office and published on a site that search engines index well is easy to find. An identical sandbox, documented in another language and uploaded as a scanned PDF, may be missed. The Index can then report a difference in visibility as a difference in readiness.
To give a fuller account of AI policy across all countries, we are expanding a process we introduced last year. Before publication, governments can submit public evidence directly, pointing our researchers to documents and programmes that may sit outside the languages and databases international research depends on.
How it works
Governments can point our researchers to public evidence for specific indicators that desk research may otherwise miss. That might be a budget line in an appropriation act, the minutes of an advisory council or the gazette entry for a data protection law. They cannot submit a score or self-assessment. We judge every document against the same criteria as evidence found through desk research. A press release does not become evidence because a ministry sends it; a programme counts when the evidence shows that it meets the indicator’s requirements, whoever found it. Accepted evidence feeds into the dataset, rankings and analysis published in the final report. Participation is optional. We research and score all countries whether or not their governments submit evidence.
Policy does not always look familiar
Desk research starts with defined indicators and an accumulated picture of what evidence for them tends to look like. That picture comes from years of work across the team and from the strategies, councils, funds and institutes governments have used before. Usually, it makes research faster and more accurate. In AI, the picture can go out of date faster than we can update it. A government may meet a familiar indicator through a new instrument such as a compute partnership, describe it in terms we have not used, or embed it in an institution with no obvious counterpart elsewhere.
Submissions show us where that picture is too narrow. They reveal policy forms our searches did not anticipate and the local language governments use to describe them. Some strengthen the evidence for an existing indicator. Others expose policy developments the framework does not yet capture.
A record that outlasts a ranking
The repository will turn those submissions into something the annual Index cannot provide: a continuous record of how AI policy develops. It will bring together countries’ own documents and terminology, preserve evidence that falls outside the current framework, and show how programmes change between one edition of the Index and the next. Policies can move from announcement to funding to implementation in a matter of months. A report published once a year captures a moment; the repository will let us follow the movement.
Governments gain something from assembling that record, too. Bringing together the budget line, the council minutes and the gazette entry requires officials to trace activity across departments and institutions. The result is a clearer account of their own work and an Index with a better chance of getting it right.
For more details on the evidence submission process for the 2026 edition of the Government AI Readiness Index, click here.
