23/07/2026
The OECD just published its first major report mapping how artificial intelligence is being used to support citizen participation in government. It draws on 50 real-world use cases from 22 countries, several of them EU member states already piloting exactly this, and asks a deceptively simple question: where does AI actually help governments listen?
It’s a question worth sitting with in Brussels. The EU has no shortage of consultation portals and citizen dialogues; what it often lacks is a way to make sense of what comes back, or a guarantee that any of it moves a decision. So before asking what AI could do for EU participation, it helps to look at a country that’s already run this experiment for over a decade.
Brazil already proved it’s possible
Long before “AI for democracy” became a policy buzzword, Brazil’s Federal Senate built a system where ordinary citizens routinely shape legislation, not just vote for the people who write it. Through the Senate’s e-Citizenship platform, Brazilians can submit a law idea in a tweet-length message, and if it gathers enough public support, it moves to committee for formal drafting. Over more than a decade, that system has collected more than 120,000 legislative ideas and 11 million votes, and helped shape 46 bills. The people behind those numbers are easy to forget, so it’s worth naming a couple. An 82-year-old retiree from Maceió called the Senate’s hotline to suggest larger font sizes on medication labels – a senator picked it up and ran with it. A citizen in Minas Gerais proposed banning internet providers from throttling services like Netflix. Five years later, net neutrality was law.
Where AI actually fits in
Brazil’s system works, but it also strains under its own success: thousands of similar ideas splinter public support, popular proposals crowd out the rest, and staff have to manually sort through a growing flood of submissions every month.
This is exactly the gap the OECD’s typology is built to address. Among the nine categories of AI tools it identifies, a few map directly onto themes the #RebootDemocracy series flagged a year ago:
- Clustering duplicate proposals, so similar ideas stop competing with each other for the same 20,000 signatures.
- Surfacing overlooked ideas, instead of letting the same three trending proposals dominate the homepage indefinitely.
- Synthesising large volumes of public comment, the way Google DeepMind’s Habermas Machine experiment showed AI can distil thousands of viewpoints into a summary that participants judge fairer than one written by a human mediator – and, crucially, without erasing minority positions.
None of this is hypothetical anymore, and it isn’t only happening outside Europe. The OECD’s case studies include the European Commission’s multilingual translation tool that let citizens submit ideas in all 24 official EU languages during the Conference on the Future of Europe, and a joint project across the Netherlands, Portugal and Spain, co-created with 90-plus participants from government, academia and civil society, building AI prototypes now being tested for national consultation platforms, including Portugal’s own Participa.gov. Belgium ran an AI-supported citizen panel on AI policy itself. The building blocks for an EU-wide approach already exist; they’re just scattered across pilots.
The test that still matters most
In the #RebootDemocracy series, Cristiano Ferri Faria, founder of the Chamber of Deputies’ Hacker Lab, offered a caveat that has stuck with us since: participation means nothing if it never touches an actual decision. That’s the line the OECD report is careful to draw too. Sense-making tools, translation, moderation, facilitation – all of it can make listening faster, cheaper and more inclusive. None of it guarantees that what’s heard changes what lawmakers do. The report is explicit that AI cannot fix the deeper disconnection between citizen input and policy outcomes; that remains a matter of political will, not technology. The OECD has now mapped 50 ways governments are learning to listen better, several of them already within EU borders. So back to the question in our title: if the EU built on these pilots and scaled them into something closer to Brazil’s decade of practice, would it actually change what the European Parliament or Commission decide – or just make consultations feel more responsive without becoming more binding? Cristiano’s test still stands as the one that separates genuine participation from theatre: did anything actually change?