Skip to content
1 min read Podcast

The Real AI Story for Down-Ballot Campaigns Isn't Deepfakes

AI is excellent at speed, scale, scope, and sophistication. It's terrible at relationships, personality, and judgment. The candidates who win will know which is which — and only automate the things that don't need to be them.

The Real AI Story for Down-Ballot Campaigns Isn't Deepfakes

The headlines about AI in politics fixate on deepfakes and disclosure rules. But the more useful story, as Becki Wright and I get into on this episode, is far more mundane — and far more relevant to the local and down-ballot candidates trying to figure out where to even start.

Becki is the founder of Proximity, a political CRM and fundraising platform built for smaller campaigns, and she brings a rare dual perspective. She's been a candidate herself — she lost a race by 11 votes running on paper maps and Google Sheets — and she spent more than a decade as a campaign manager, finance director, and consultant on races from school board to U.S. Senate. That experience convinced her campaign tech was lagging badly behind the tools available everywhere else.

Most campaigns using AI today reach for general-purpose tools like ChatGPT, Claude, or Gemini. Those models help with opposition research, ad copy, and speeches, but they lack the local context that matters — the district, the history, the constituents. Worse, their usage policies are often openly hostile to political work. That's why Becki argues purpose-built AI matters: the boring reality of campaigning is yard signs and voter contact, not dark arts.

She's candid about the failure modes. Off-the-shelf chatbots suffer from sycophancy (telling candidates their ideas are brilliant), an averaging effect that buries the important 10%, and hallucination. Her fix is a layered approach — using rules and a second tool to check the first — and keeping a human in the loop.

We also dig into voter trust. New polling from me and the AAPC shows a strange dynamic: obvious AI disclaimers can actually lower trust, because anything that smells inauthentic undermines the connection campaigns depend on. Becki's rule is simple: use the candidate's real words and face, then let AI handle repurposing and volume.

My takeaway is a discipline question. AI is excellent at speed, scale, scope, and sophistication. It's terrible at relationships, personality, and judgment. The candidates who win will know which is which — and only automate the things that don't need to be them.