Voters Are Worried About AI in Politics. Here's What Good AI Campaigning Actually Looks Like

Key points

• AI deepfakes, fake videos or audio clips generated by artificial intelligence, are already being used to spread misinformation in US elections.
• Campaigns in Japan and Scotland have tested AI tools that help candidates gather and understand voter concerns at scale.
• US academic and private institutions are exploring similar listening-first approaches, but mainstream adoption in American politics remains rare.
• Experts argue AI could make political engagement more inclusive and policy more responsive, if candidates choose to use it that way.

AI is already showing up in US elections. The problem is how.

Deepfakes, fake videos and voice clips generated by artificial intelligence, are spreading lies about candidates. Official accounts are posting what critics are calling "slopaganda", low-effort, AI-generated content designed to flood feeds rather than inform voters. And the public, understandably, is anxious.

But buried under all that noise is a question worth asking: what if campaigns used AI to listen instead of just broadcast?

What does good AI in politics actually look like?

It looks less like a robot giving speeches and more like a very patient aide reading thousands of emails. In Japan and Scotland, political groups have experimented with AI tools that sift through huge volumes of constituent messages, public comments and survey responses, then summarise what voters actually care about. That lets a candidate or their team spot patterns a human staffer would miss after three days of reading.

Scotland's example is particularly instructive. Civic tech groups there used AI to process public input during policy consultations, helping officials understand which concerns came up repeatedly across different communities. The goal was not to replace human judgment. It was to make sure a pensioner in a rural town and a young renter in Glasgow both got heard, not just the loudest voices in the room.

Japan has seen similar experiments, with AI tools helping local candidates analyse constituent feedback faster than any team of humans could manage.

Why isn't this happening in the US?

Vanishingly few US campaigns are doing this, as The Guardian AI noted in its original coverage. The incentives point the other way. Targeted advertising, outrage content and deepfakes are cheap to produce and demonstrably good at grabbing attention, even if they corrode trust over time.

US universities and some private technology firms are building listening-focused tools, but moving them from a research paper into a real campaign is a different challenge entirely.

The gap is not a technical one. The tools exist. It is a political will problem.

What does this mean for ordinary voters?

Right now, not much changes at the ballot box. But the direction matters. If campaigns start competing on who listens best rather than who shouts loudest, the technology that currently fuels misinformation could, in a different set of hands, help produce policy that actually reflects what people need.

For now, the practical advice is simple: treat AI-generated political content with the same scepticism you would a cold call asking for your bank details. If a video or quote from a politician surprises you, look for a second source before sharing it.

The tools to do politics better already exist. The question is whether anyone running for office decides to pick them up.

Common questions

Can AI really understand what voters want?

AI text-analysis tools can identify common themes across thousands of messages far faster than a human team, but they cannot judge nuance or lived experience on their own. They work best as a filter that surfaces patterns for human decision-makers to act on.

Are AI listening tools being used in the UK or Europe right now?

Small-scale civic tech projects in Scotland and parts of Europe have tested these approaches in policy consultations. Widespread use by mainstream political parties has not been confirmed.