Is it legal to use AI in a political campaign?

Every campaign manager who hears “AI voter outreach” asks the same first question, usually with their lawyer on the line: is this even legal? The honest answer is yes, with rules. As of 2026 there is no federal ban on using AI to answer voters, and the regulators who looked at it decided the laws you already follow still apply. The mistake is not using AI. The mistake is using it in a way that breaks a rule that was already on the books.

This is a guide, not legal advice. Election law is federal, state, and local all at once, and it moves. Confirm anything here with your campaign counsel and the current FEC, FCC, and state rules before you deploy.

Is it legal to use an AI chatbot for a political campaign?

Yes. In 2024 the Federal Election Commission weighed whether to write new, AI-specific rules and declined, concluding that its existing regulation on fraudulent misrepresentation is technology-neutral and already covers AI. In plain terms: the FEC did not create a new “AI is illegal” rule, and it did not carve out an exception either. The same laws that governed a mailer or a phone bank govern an AI that answers a voter.

So the real question is not “is AI allowed.” It is “does this specific use follow the rules that already exist.” An AI that answers your own voters, in your own campaign’s voice, with a clear disclosure, is on solid ground. An AI that impersonates another candidate, hides who paid for it, or dials people who never opted in is not — and it would have been illegal with a human doing it too.

The rules that actually govern AI on a campaign

Four bodies of law do most of the work. None of them is about AI specifically; all of them apply to it.

What you have to disclose

Three disclosures cover most situations. First, the “Paid for by” attribution that already goes on your ads goes on every AI surface too: the website assistant, the text messages, the outbound call script. Second, get consent before you text or call a mobile number; opt-in is not optional under the TCPA. Third, where your state has a synthetic-media or AI-disclosure law, label AI-generated audio and video in ads. When in doubt, disclose more, not less. Transparency is cheap, and it is the single best defense if anyone ever questions the outreach.

AI voice calls: the “Ashley” line and the FCC

AI on the phone is where campaigns get nervous, for good reason. In 2023 a congressional campaign ran an AI phone canvasser named “Ashley” that openly told voters it was an AI: disclosed, consented, legal. A few months later, someone sent AI-generated robocalls faking a sitting president’s voice to suppress turnout in a primary. The FCC responded by ruling that AI-generated voices in robocalls count as “artificial or prerecorded” under the TCPA, which means they require prior consent and carry real penalties.

The line between those two examples is the whole lesson. A disclosed AI that a voter agreed to hear from is fine. A voice clone pretending to be a real person, blasted to people who never opted in, is illegal — and the technology being “AI” is not what makes it illegal. Deception and lack of consent do.

A chatbot trained on the candidate’s positions

The useful version of campaign AI is not a generic bot bolted onto a website. It is one approved brain that knows the candidate’s actual positions, answers on message across every channel, and refuses to freelance. A voter asks about a stance in web chat at midnight, texts a follow-up the next day, then calls the office, and gets the same answer every time, grounded in positions the campaign approved, because there is one source behind all of it.

That grounding is also the guardrail. The brain answers from approved material and hard-blocks the topics you never want it near. When a question needs a human — a press inquiry, a sensitive constituent issue, a legal matter — it hands off to your team instead of guessing. Every conversation is logged, so there is a record of what was said, on which channel, and when.

AI for voter outreach and GOTV

The reason to do any of this is capacity. A campaign has a fixed number of volunteers and an unlimited number of questions. An AI that instantly answers “where do I vote,” “what is the candidate’s position on this,” and “how do I get a mail ballot,” 24/7, in the candidate’s voice, across web chat, SMS, and a phone line, frees your people for the conversations that actually need a human: persuasion, volunteer coordination, and get-out-the-vote to the voters who move the margin. It is not a replacement for field. It is what lets field spend its hours where they count.

How to deploy AI on a campaign without getting burned

The compliant version of campaign AI is a set of controls, not a vibe. Before you turn anything on, make sure it does all of this:

• Inbound and opt-in only — never message or call a number that did not consent
• “Paid for by” disclosure on every surface: the widget, the texts, the call
• Never impersonates anyone — it speaks only as your campaign
• Hard-blocked topics it will not touch, set by you
• Grounded in the candidate’s approved positions, not the open internet
• A full audit log of every conversation, on every channel
• Clean handoff to a human the moment one is needed
• A kill switch that stops it instantly

Get those right and AI stops being a legal risk and becomes what it should be: more voters answered, on message, at a fraction of the cost, with a record to prove every word of it.