Australian MPs lead English-speaking world in AI-generated speeches

Three years ago, Liberal MP Aaron Violi made a choice that seemed almost quaint in retrospect: he publicly announced that he had used artificial intelligence to write a parliamentary speech. It was a declaration meant to demonstrate the technology's capabilities, a kind of controlled experiment in transparency. Today, that moment reads like the opening chapter of a much larger story. Australian politicians have become the heaviest users of AI-generated speeches in the English-speaking world, according to analysis published in The Economist, outpacing even Canada and running at triple the rate of British MPs.

The numbers are striking. Between September 2025 and September 2026, researchers sampled speeches delivered in Australia's House of Representatives and flagged more than one in five for AI-generated or AI-assisted content. In the Senate, the figure dropped slightly to just under 15 per cent. These are not marginal practices or isolated experiments. They represent a shift in how elected representatives are constructing the words they deliver to their peers and, by extension, to the public record.

The trend has surfaced a tension that sits at the heart of democratic representation. Violi himself, now the opposition spokesman for digital economy and cybersecurity, acknowledges the utility of the tool while warning against its careless deployment. He encourages his own staff to use AI, but only within strict guardrails: facts must be verified against source documents, confidential information must never be entered into the system, and the output must never be used verbatim. "If an MP is putting a prompt into AI and then using the speech word-for-word without checking the source documents, well, I think at some point they're going to get themselves in trouble," he said. He notes that AI, by its nature, produces the statistical average of existing content—safe, predictable, stripped of the distinctive voice that might actually move an audience. His own most effective parliamentary moments, he says, come from speaking off the cuff with only dot points as a guide.

Speechwriters who have worked for former prime ministers offer competing perspectives on what is being lost and gained. James Button, who wrote speeches for Kevin Rudd and previously worked as a journalist, frames the problem in human terms. "The thing about a speech is that it's a very human event, right? It's a person standing up in a room, speaking to others, and seeking a connection with those people," he said. "You add AI into that, and we are absolutely not getting the thoughts of the human being, the politician." Another former prime ministerial speechwriter, speaking anonymously, sees the efficiency gains as real and substantial—what might take days to craft by hand can now be generated in minutes and then refined. But this convenience carries a cost. Those who use the tools well will sharpen their skills; those who simply substitute AI for the talking points they once recycled will have merely traded one form of artificial thinking for another.

Independent Senator David Pocock has published his own guidelines for AI use and is pushing for something Parliament currently lacks: transparency requirements. He wants MPs to disclose when they have used the technology and to justify that use to their constituents. His concern is not that AI exists, but that its proliferation is producing a kind of homogenization—politicians, as he puts it, "bludgeoning people with platitudes." The human connection between representative and constituent, he argues, cannot be sacrificed on the altar of efficiency. "As an elected representative, you're elected to represent, not to read out AI speech. It comes down to trust. I think Australians would expect you to be using it to become more efficient. But you can't lose that human touch."

Speaker Milton Dick has confirmed that Parliament has no formal rules governing how members use AI in their speeches. Members remain responsible for what they say, he noted, and they carry an overriding duty to the national interest and to their constituents' interests. But responsibility and rules are not the same thing. The absence of formal guidance means that the decision to use AI—and how to use it—remains entirely with individual MPs. Some will approach it with the caution Violi describes. Others will not. The question of whether Parliament should establish clearer expectations around disclosure and use remains unresolved, even as the practice itself has become commonplace.