Steve Keen predicts AI will not move votes in 2028, and the reasoning is worth examining

Economist Steve Keen has placed a specific bet: AI will not be a major issue in the 2028 US presidential election. His reasoning rests on a claim about mass experience. Most Americans, he argues, encountered AI at its worst, concluded it was not worth their attention, and returned to the rest of their lives. The people who stayed, who watched the technology improve and integrated it into their work, are in his view a minority. The majority never came back.

That framing is falsifiable. Either AI becomes a live electoral issue by 2028 or it does not. Keen’s reasoning is worth spelling out precisely because the conditions he describes could change well before the next election cycle closes.

The core of Keen’s argument is a gap between the AI-fluent and everyone else. “The overwhelming majority of people don’t have that experience,” he notes, “and most people’s biggest exposure to AI is they tried an early model two years ago, it was totally inaccurate, and they went oh this is a bit and moved on with their lives.” The logic follows: voters who never developed a functional relationship with the technology have no strong stake in how it develops. No stake, no vote-moving salience.

The overwhelming majority of people don't have that experience and most people's biggest exposure to AI is they tried an early model two years ago it was totally inaccurate and they went oh this is a bit and moved on with their lives. Steve Keen

That portrait of a largely disengaged public is not implausible. Keen’s prediction depends on that pattern holding. If voters continue treating AI as peripheral, something the professional class obsesses over while everyone else has more pressing concerns, then the issue never organizes into the kind of wedge that shapes electoral outcomes. Three years is a long time for a new topic to either mature into a genuine grievance or fade into background noise.

The structural risk to Keen’s thesis is not attitude change but circumstance change. The voters who tried a bad model and moved on did so when AI affected very little of what they actually did. If AI-driven automation reaches their sector, or their employer announces restructuring tied explicitly to the technology, or energy costs rise in ways that can be traced to data-center expansion, the prior experience of a clunky chatbot stops being the relevant reference point. Keen’s prediction requires not just that most people tuned out once, but that nothing between now and November 2028 gives them a compelling reason to tune back in.

Electoral salience requires more than ambient anxiety. It requires anxiety to organize around candidates, choices, and platforms in ways that actually shift vote totals. That organizational step has not happened yet on this issue, and Keen’s broader point, that the majority of the electorate has a thin and negative first impression of the technology rather than a developed opinion, remains a reasonable description of where things stand. The question his prediction leaves open is whether that thin impression is stable or simply dormant.

Keen is not predicting that AI will be unimportant. He is predicting that it will be unimportant electorally, specifically in 2028, in the United States. That is a narrower claim and a more defensible one. Issues can reshape economies and workplaces without ever organizing into decisive voting blocs, particularly when their effects distribute unevenly across sectors and regions. His bet is that AI falls into that category, at least on the 2028 timeline. Whether the disruptions the technology is already producing diffuse quietly enough to stay off the ballot, or concentrate sharply enough to force their way onto it, is the test his prediction sets up and that only November 2028 can settle.