Skip to content
Citizen Infrastructure

Takahiro Anno's AI-augmented Tokyo campaign (2024)

In the July 2024 Tokyo gubernatorial election, AI engineer Takahiro Anno (33), an independent newcomer with no political machine, ran a campaign built around broad listening: a manifesto on GitHub that anyone could propose changes to, an AI avatar that answered voters’ questions, and opinion-clustering software to surface what the public actually wanted. In a field of a record 56 candidates he won roughly 150,000 votes — and the method, more than the result, is what spread.

At a glance

Section titled “At a glance”
Process typeAI-augmented election campaign / broad listening
WhenJuly 2024 (Tokyo gubernatorial election)
WhereTokyo, Japan
WhoTakahiro Anno — AI engineer & science-fiction writer — and a volunteer team (later “Team Mirai”)
ToolsGitHub manifesto (open to pull requests), an AI avatar, Polis + “Talk to the City” for clustering opinion
Result~150,000 votes (5th place); the playbook adopted far beyond the campaign

What happened

Section titled “What happened”

Anno read the Plurality book and, about a month before the election, decided to run — not really expecting to win, but to demonstrate a different way to campaign (the Taiwan “demonstration, not protest” move). Instead of broadcasting a fixed platform, he inverted it into listening:

  • His manifesto lived on GitHub, and anyone could open a “pull request” suggesting a change; the campaign reviewed and accepted dozens of them, so the platform improved through the campaign.
  • An AI avatar of Anno let any member of the public ask about his policies and make suggestions at any hour, around the clock.
  • “Talk to the City” and Polis-style tools clustered the incoming opinions and reflected back what people thought, so suggestions could be turned into concrete platform edits.

No person under 40, with no formal political organisation, who had never run before, had previously broken 50,000 votes at that scale in Japan.

Inputs → outputs

Section titled “Inputs → outputs”

In: thousands of citizens’ questions and suggestions, gathered through the avatar, the GitHub repo, and opinion-clustering tools.

Out: a continuously revised, open-source manifesto (dozens of public contributions merged), and ~150,000 votes — far beyond any comparable newcomer, placing fifth.

Impact

Section titled “Impact”

The campaign’s real result was legitimising the method. Anno went on to found Team Mirai (“Team Future”), a political group built around e-democracy and “broad listening” tools; in July 2025 it secured its first seat in Japan’s House of Councillors, with Anno elected. The approach drew international attention as a rare example of AI used to strengthen rather than undermine an election — generative AI pointed at listening to constituents instead of flooding them with synthetic content.

How it went

Section titled “How it went”

A genuine proof of concept that AI-augmented conversation networks can run inside a real, high-stakes election, not just a government consultation. The honest caveats: fifth place is influence, not office; the tooling demands technical capacity most candidates lack; and Team Mirai has been described as technocratic, so whether “broad listening” scales into durable governance is still being tested. As a demonstration that a campaign can be built on listening rather than broadcasting, though, it landed.

Sources

Section titled “Sources”