When AI Can Explain Government but Cannot Reach It
Artificial intelligence is becoming an intermediary between governments and the people and businesses that depend on public services. In their July 2026 World Bank preprint, RADAR: Readiness for AI Discovery and Agentic Reach, Luke Jordan, Tiago C. Peixoto, and Manuel Ramos-Maqueda examine a question that conventional digital-government rankings largely overlook: can AI systems not only explain government services, but actually reach the official digital channels needed to use them?
The authors develop the RADAR framework across 166 countries and distinguish two separate dimensions of AI readiness. “Informational legibility” measures whether large language models can provide specific, verifiable, officially sourced guidance about government services. “Agent operability” measures whether automated agents can navigate government websites and reach the relevant service entry point. The chat component tested 20 service scenarios across four frontier models, while operability was tested through both document-object-model navigation and visual browser interaction. That distinction is central because a government can be well represented in AI-generated explanations while remaining operationally difficult for an automated agent to access.
The empirical result is striking. Across all 166 countries, informational legibility exceeded average agent operability, with a mean guidance-to-reach gap of 2.37 points on a ten-point scale. The gap also did not disappear with higher national income. Estonia ranked first overall, while the United States ranked fortieth. More importantly, the study finds that traditional digital-government strength does not necessarily translate into agent accessibility. Sophisticated portals may still frustrate automated navigation because of client-side rendering, unstable URLs, weak deep-linking, cross-domain complexity, or aggressive bot controls.
Jordan, Peixoto, and Ramos-Maqueda therefore argue for what they call “sovereign legibility”: deliberate government action to make authoritative content discoverable, structured, and interpretable by AI systems. Their recommendations include canonical service pages, stable URLs, structured metadata, machine-readable content, calibrated bot controls, documented APIs, and standardized protocol layers such as Model Context Protocol interfaces. The authors point to the U.S. Government Publishing Office’s public GovInfo MCP server as an early example of government information being exposed through an officially supported machine-facing channel.
The study also appropriately cautions against treating all friction as a defect. Authentication, payment, identity verification, and other legally consequential transactions may require barriers that protect security and authorization. The authors also acknowledge methodological limitations, including English-language chat prompts, use of verifiability rather than independently established ground-truth accuracy, and an agent-operability test centered on a single public service. RADAR is therefore most useful diagnostically: governments should distinguish poor information architecture from intentional controls rather than simply maximizing automated access.
For federal contractors, the broader lesson is increasingly relevant. As agencies, offerors, and advisors use AI to retrieve, interpret, and act on government information, the quality of the underlying structure matters. Procurement information that is fragmented, buried in attachments, inconsistently labeled, or difficult to trace to authoritative sources creates an analogous “legibility” problem. Contractors should therefore treat structured compliance analysis and authoritative-source traceability as part of proposal discipline, not merely as administrative housekeeping.
Recommended FedContractPros.com Tool
The Section L/M Compliance Crosswalk is the strongest fit. RADAR’s core insight is that information becomes more reliable when it is structured, traceable, and connected to an authoritative source. The same principle applies to federal proposals: the Crosswalk helps contractors systematically map solicitation instructions and evaluation criteria to proposal content, reducing the risk that requirements become fragmented, overlooked, or inadequately addressed.
Disclaimer
This post is provided for educational and informational purposes only and does not constitute legal, procurement, technology, or professional advice. It summarizes and comments on research by Luke Jordan, Tiago C. Peixoto, Manuel Ramos-Maqueda, and their contributors. FedContractPros.com is not affiliated with or endorsed by the World Bank or the authors.