AI Can't Rescue a Channel Voters Already Hate
One firm sent two and a half million political text messages this year to produce somewhere between twenty and thirty thousand conversations. Another is racing to work with more than a hundred campaigns before November. The pitch behind the surge is that generative AI has finally made political texting smart — bots trained to sound like the candidate, answering questions about tax policy or school funding in thirty seconds, in any language, at a scale no volunteer operation could match.
It's a real capability, and it's being sold hard. But it answers the wrong question. The problem with political texting was never that the messages weren't clever enough. It's that voters stopped trusting the channel years ago. A smarter bot doesn't fix that. It just makes the thing people already resent a little more convincing.
The Channel Was Broken Before the Bots Arrived
A recent NPR report put a face on the fatigue: a voter in Cleveland who started getting as many as five political texts a day, clickbait-y in tone, impossible to trace to a sender. "I really have no idea who's on the other side of that," she said. That reaction isn't an edge case. It's what happens to any channel that gets abused faster than it gets respected.
The clearest line in the piece came from someone whose job is helping campaigns adopt exactly this kind of technology. AI, she said, "is not best used to rescue channels that people already hate. It's best used to find new ways to reach people." That's the whole argument in a sentence. Pointing a large language model at a channel voters have been trained to swipe away doesn't restore the trust that made the channel work in the first place. It spends down what little is left.
We've watched the same movie play out in the inbox. An office or a campaign builds a list, then blasts it with undifferentiated volume until open rates collapse and the sends quietly start landing in spam. The instinct is always to reach for a new tactic — a new tool, a new channel, now a new bot. The tactic is rarely the problem. The relationship is.
The Fix Isn't a Better Bot. It's Earning the Right to Be There.
There's a phrase we come back to constantly: earn the right to someone's inbox before you ask anything of them. It applies just as cleanly to a phone. A message earns a response when it's genuinely useful to the person receiving it, and it burns goodwill when it isn't. No amount of conversational fluency changes that math. A bot that replies in thirty seconds is still an interruption if the voter never wanted the conversation.
This is where the discipline actually lives. On one flagship program we run, email open rates average 67.8 percent — roughly three times the political and advocacy norm — with peaks above 80 percent on individual sends. That didn't come from sending more or writing cleverer copy. It came from being ruthless about relevance: agriculture content went to farmers and ranchers, a dedicated note about school funding went to parents rather than getting buried as one bullet in a generic newsletter. People open what's meant for them. They ignore, and eventually resent, what obviously isn't.
The same principle governs new contacts. Rather than pull a fresh name straight into an ask, we warm it first — a sequence designed to be worth opening before it's ever worth anything to us. On that program, those warming sequences convert new contacts at a 31 percent clip. The channel doesn't matter as much as the sequence of trust behind it.
Where AI Actually Belongs
None of this is an argument against texting, and it isn't an argument against AI. Texting still does something no other channel does — it reaches voters directly, on a device they check constantly, without competing against a social platform's algorithm for attention. It isn't going away, and campaigns are right to use it.
The question is what you point the intelligence at. Used well, AI is a segmentation and listening engine — a way to understand what an audience actually cares about so the next message is more relevant, not just more automated. Used poorly, it's a way to manufacture the appearance of a personal conversation with someone who never agreed to have one. The first compounds trust over a cycle. The second spends it, and does so faster the better the technology gets.
Republican campaigns, by most accounts, are adopting these tools faster than the other side right now. That's an advantage worth having — but only if the tools get aimed at the right target. The edge won't belong to whoever deploys the most convincing bot. It'll belong to whoever built an audience that already trusted them before the bot ever sent a word.
An engaged audience is worth far more than a large one, and it's worth more still than a heavily automated one. That was true when the channel was email, it's true now that the channel is text, and it will be true for whatever comes next. The technology keeps changing. The reason people open a message hasn't.