AI political campaigns are learning to run themselves

We may soon find out what democracy looks like when humans aren’t included.

So, I took two months off. Not for a fancy getaway or anything cool like that. Rather, life intervened, as it tends to do from time to time. Health issues pulled me offline and sent my life into a tailspin, at least temporarily. For a while, the constant scroll of chaos and doom faded into background noise. And for the first time in years, I took a break from dissecting the latest viral disinformation and outrage bait.

When I came back, I expected to find the same disinformation playbook humming along — the usual blend of outrage cycles, coordinated inauthentic behavior, and platform manipulation. And that was, indeed, waiting for me. But what I also found was that while I was gone, AI started running for office.* Well, sort of.

*Admittedly not all of these events took place during my absence, but stepping back is what allowed me to see them as part of a larger and unnerving pattern.

The Campaigns That Ran Themselves

In New York, a congressional campaign released a fully AI-generated ad — synthetic voice, algorithmic script, composited imagery — everything but the handshake. It was eerily plausible and completely detached from any camera or crew. This wasn’t a deepfake of a rival, though; it was a campaign…making itself.

In Wyoming, a man named Victor Miller made headlines for filing paperwork to run for mayor of Cheyenne — on behalf of an AI bot called VIC, short for “Virtual Integrated Citizen.” Miller’s pitch was simple and absurd: he’d serve as the bot’s physical proxy while the algorithm made the actual decisions. When asked who voters would be electing, he answered, “VIC. I’m just the vessel.” The Secretary of State’s office, not yet coded for existential paradoxes, rejected the bid.

Across the Atlantic, British lawyer Andrew Gray ran for Parliament as an “AI-assisted candidate,” using a chatbot trained on local input to crowdsource his platform. The system, nicknamed Polis, synthesized thousands of citizen comments into policy positions. Gray didn’t win, but he proved a point: it’s now possible to outsource empathy. Or at least something that looks like it.

And then came the ads — some styled as parodies and prototypes, but also campaign ads themselves. Ok

Just this month, former New York governor Andrew Cuomo released an AI-generated ad showing himself as a window washer, theater worker, and nurse — a montage of working-class avatars, stitched together to (re)build his political brand. The spot was supposed to convey humility. Instead, it felt like something out of a focus-group hallucination: a man trying to prove he’s human by distinctly non-human means.

From Deepfakes to Deep Politics

None of this should be surprising. Generative AI was always going to crash into the political process. But what’s happening now isn’t the rise of the “AI politician” — it’s the automation of persuasion.

Campaigns have always been about narrative engineering. What AI has done is remove the last bottleneck: time. A small team can now generate hundreds of localized ads in a day, tailor each message to a psychographic profile, and even simulate the cadence and emotional tone most likely to move you.

In other words, the campaign doesn’t need to know you anymore; it can model you.

On its face, the logic is almost irresistible. Why spend weeks on focus groups when you can just generate synthetic voters? Why test slogans when you can run infinite A/B scenarios overnight? Every campaign manager in America is quietly tinkering with this, even if they won’t admit it on record.

The line between propaganda and personalization has dissolved — and in its place, we’re left with something new: programmable persuasion.

The AI Candidate Is a Mirror

To be clear, the AI candidate is not here — not yet, at least. But symbolically, it’s already arrived. When a campaign like Cuomo’s uses AI to simulate human struggle, it’s not selling policy; it’s selling relatability. When Victor Miller says his chatbot should govern, he’s not trolling — he’s voicing a growing cynicism about human governance itself.

The message beneath these stunts is corrosive but effective: maybe the machines would do it better. Some of the most esteemed minds in AI have made the same argument when suggesting that AI replacing the human race may not be such a bad thing after all. That’s where this line of reasoning ultimately leads.

And that’s how technology eats politics — through resignation, not revolution.

For decades, campaigns were mediated by people: speechwriters, pollsters, journalists, etc. AI is cutting out those middle layers. A candidate’s digital team can now fabricate an entire rhetorical ecosystem — posts, quotes, reactions, rebuttals — and seed it across platforms before dawn. It’s like Russia’s 2016 campaign, except automated and on the digital equivalent of steroids.

It’s not that voters can’t tell real from fake — it’s that they don’t have time to care. The constant churn of generated content erodes discernment. Everything feels equally plausible and equally disposable.

That’s the new terrain of democracy: instead of truth versus lies, it’s attention versus entropy.

The Machine Campaign

There’s an old phrase in political science — “the permanent campaign.” It used to mean the endless fundraising and messaging between elections. Now, it’s literal. AI tools don’t need sleep or budgets. They can microtarget perpetually, refining their messaging loops long after election day.

Imagine a campaign that never turns off.

A perpetual presence, adaptive, omnipresent, algorithmically fine-tuning its persona to whatever you need it to be.

That’s not 2030. That could be the next election cycle.

In 2024, we got glimpses: AI-generated robocalls impersonating President Biden, deepfake videos of local officials, synthetic influencers amplifying political narratives in a long list of languages. These operations aren’t domestic; they’re transnational, blending state actors with mercenary firms and AI disinformation networks that evolve like malware.

The U.S. intelligence community calls this “AI-enabled influence operations.” It’s a bureaucratic understatement for something much darker: the industrialization of manipulation.

The Post-Truth Economy

What worries me most isn’t the technology — it’s the economy of indifference it feeds. Each time a campaign normalizes AI fakery, it resets the moral bar lower. “Everyone’s doing it” becomes the defense. Transparency becomes optional.

Meanwhile, voters are conditioned to doubt everything. Every clip, every quote, every stumble becomes questionable. That’s not healthy skepticism; it’s epistemic exhaustion — the point where the difference between truth and illusion stops mattering.

And that exhaustion is the perfect environment for authoritarianism. When nothing feels authentic, the strongest voice wins by default.

The Human Element

When I stepped away for those two months, I expected to lose touch. Instead, I came back and realized the opposite: it’s the world that’s losing touch — with reality, with embodiment, and with the limits that make democracy human.

The irony of my absence is that I took a break because I’m human. Fallible, biological, slow. The same qualities our political system depends on — deliberation, empathy, accountability — are precisely what AI politics erodes.

If democracy is, at its heart, a conversation among flawed beings, what happens when that conversation is simulated?

What happens when the candidate never sleeps, never hesitates, never reveals a crack in the voice or a pause before the answer?

We’ve always said we wanted our politicians to be the authentic. Yet now, we’re building machines to fake it better. And research suggests that these machines will almost certainly be damaging to democratic societies, leading to more inflammatory and extreme political campaigning.

The CAPTCHA Moment

There’s a phrase I can’t shake: Democracy doesn’t have a CAPTCHA.

We have no way to prove, in any meaningful sense, that our politics are still human. Platforms are half-heartedly adding watermarking tools and AI disclosure labels, but enforcement is spotty, and the incentives point the other way. Synthetic content is cheaper, faster, and performs better in the algorithmic feed.

Unless journalists, watchdogs, and yes — voters — start demanding proof of reality, we’ll wake up in a polity where authenticity itself becomes a partisan issue.

I came back from my unplanned hiatus to find the boundaries of the real collapsing faster than the fact-checkers can type. The campaigns have become code. The candidates are starting to blur.

And in that blur lies the future of persuasion — programmable, persuasive, and profoundly post-human.

Author’s Note

I’m back. Still human. For now.

Updates are coming soon, so stay tuned to Weaponized Spaces so you don’t miss anything!