Updated October 2026
A single AI-generated image can run in a dozen states in an afternoon. The disclosure rules it meets along the way are not the same.
Key takeaway: No single nationwide AI disclaimer covers every political-ad requirement. States define synthetic content, deceptive depictions, covered communications and required disclosures differently. Florida, Colorado, Michigan and Wisconsin show how much the rules vary, and every state where an ad will run needs a current legal check.
This is general information, not legal advice. Check current rules with your campaign’s counsel.
Why a national AI-ad rulebook is difficult
Campaign creatives routinely cross state lines through fundraising lists, streaming video, social feeds and email. A graphic that is an ordinary campaign illustration in one context may be regulated synthetic election content in another. The operative questions are what the content depicts, whether the depiction is materially false, what the sponsor intended, which election is implicated, when the message appears and what the actual statute says.
It is unsafe to assume “AI used” always triggers a disclaimer, or that “just a joke” always creates an exemption. Some statutes focus on deepfakes of real people. Some focus on deceptive election information. Some address express advocacy, paid advertisements or a pre-election window. Court rulings can also affect how a particular statute is enforced.
Platform labeling and state election law are separate layers. Google requires election advertisers to mark synthetic or altered content through a campaign setting, even when a label also appears in the ad itself. Using a platform label does not automatically satisfy a state’s wording, placement, duration or accessibility requirements.
Four statutory examples worth reading first
Florida — §106.145. Florida’s statute covers specified political advertisements, electioneering communications or other political ads containing generative-AI content that appears to show a real person taking an action that did not occur, when created with an intent to injure a candidate or deceive about a ballot issue. It prescribes a prominent disclaimer, “Created in whole or in part with the use of generative artificial intelligence (AI),” with size and duration rules for print, video, internet and audio. The precise statutory trigger matters: the law is not a blanket label requirement for every computer-assisted image.
Colorado — HB24-1147. Colorado’s candidate deepfake law took effect July 1, 2024. It prohibits distributing a communication about a candidate that includes an undisclosed or improperly disclosed deepfake, when done with knowledge or reckless disregard of its falsity. The required statement reads: “This (image/audio/video/multimedia) has been edited and depicts speech or conduct that falsely appears to be authentic or truthful.” The act also lets the depicted candidate sue. Review the definitions, exceptions and placement rules in full before distribution.
Michigan — MCL 169.259 (2023 Public Act 263). Michigan requires qualified political advertisements to state, in a clear and conspicuous manner, that they were generated in whole or substantially by artificial intelligence, with format rules for graphic, audio and video communications. Its definition and method of disclosure differ from Florida’s intent-based provision.
Wisconsin — 2023 Act 123. According to the Wisconsin Legislative Council, Act 123 requires a disclosure on audio or video communications that contain express advocacy or issue advocacy, or that support or oppose a referendum, when they include “synthetic media” substantially produced by generative AI. The Wisconsin Ethics Commission enforces it. Check the current statutory text before using a Wisconsin disclaimer.
These examples show why adopting the longest disclaimer found in one state is not a substitute for actual compliance analysis.
Verified examples at a glance
| State | Law | What triggers it | Disclosure |
|---|---|---|---|
| Florida | §106.145 | Generative-AI depiction of a real person doing something that did not occur, made with intent to injure a candidate or deceive on a ballot issue | Prescribed AI statement with size and duration rules |
| Colorado | HB24-1147 | Undisclosed deepfake about a candidate, distributed with knowledge or reckless disregard of falsity | Prescribed “has been edited” statement, also in metadata |
| Michigan | MCL 169.259 | Qualified political ad generated in whole or substantially by AI | Clear and conspicuous AI statement with format rules |
| Wisconsin | 2023 Act 123 | Audio or video political ad containing synthetic media | Required disclosure; Ethics Commission enforces |
Rules in other states are changing, so check the current statute and guidance from your state election office before any ad runs there. A chart built from headlines about bills that never became law is not a compliance tool.
Four things campaign creatives need to document
1. Source material and the change made
Save the original photograph, audio or video; the approved version; and a short record of what was altered. Distinguish cropping and ordinary color correction from fabricating an action, quote or voice. A source archive lets staff answer factual questions quickly.
2. Who is depicted and what is being claimed
If the asset depicts a real candidate doing something that did not happen, the risk is different from a clearly fictional illustration. An AI-generated background may also introduce inaccurate signs, ballots or people. Review the complete composition, not only the main face.
3. Placement and timing
A disclosure may need to be audible, visible, prominent or present for a particular duration. A small label that disappears in a mobile crop may fail even where the wording is accurate. If a state law applies only within a set period before an election, confirm those dates instead of guessing.
4. Platform disclosure is not statutory disclosure
Meta, Google and other channels publish their own synthetic-content policies. Entering the correct platform metadata and including the legally necessary on-creative disclosure are separate tasks. Obtain approvals on both tracks.
Our digital ads page covers how we handle platform approvals.
A safe review workflow for campaign assets
Start with the original source, identify every generated element, evaluate whether it depicts real people or events inaccurately, and decide whether the asset is necessary. If the message works with an authentic photo or an ordinary graphic, avoid creating a needless synthetic-depiction issue.
Next, list all distribution jurisdictions and election dates. Counsel should apply the relevant law, not a generic checklist. The creative team then places required disclosures in versions suitable for square social graphics, vertical video, horizontal TV and audio-only placements.
Have a second reviewer inspect the compressed mobile asset. A legally accurate disclaimer that is unreadable to actual viewers is an avoidable problem. Freeze the approved version and maintain an asset register showing where it ran and when.
Build that review into production for TV and radio, digital ads, P2P texting, automated calling and live calling scripts alike, with counsel making the final legal call.
AI creative review checklist
- Preserve original source files and alteration notes.
- Determine whether a real person’s speech or conduct is falsely depicted.
- Record intended distribution states and election dates.
- Check enacted statute text, not only pending bills.
- Check court orders and current enforceability.
- Obtain state-specific disclaimer language and placement instructions.
- Complete each platform’s synthetic-media declaration separately.
- Review mobile readability and audio intelligibility.
- Approve the final rendered asset, not just the design file.
- Record version, sponsor, dates and placements for later audit.
AI disclosure questions
Does every AI-assisted political graphic need a label? Not necessarily under every statute. Triggers depend on content, deception, intent and jurisdiction. Platform policies may separately require disclosures.
Is a platform’s AI label enough? Not automatically. State law may prescribe different wording or placement.
Can campaigns rely on a 50-state internet chart? Only after checking its date, primary-law citations and litigation updates. Any state you cannot confirm needs counsel’s review.
Does satire receive an automatic exemption? No universal exemption exists. Read each statute’s own exceptions before relying on one.
Statutes and policies cited
- Florida Statutes §106.145
- Colorado General Assembly — HB24-1147
- Michigan Legislature — MCL §169.259
- Wisconsin Legislative Council — AI Election-Law Materials
- Google — July 2026 update on AI labels in election ads
- U.S. Election Assistance Commission — Find Your State Election Office
Authentic sources, clear disclosures and careful review protect the message without needless creative risk.
Using AI in campaign creative?
We can help you build disclosure review into production, so the asset that runs is the one counsel approved. That same process carries across digital ads, P2P texting, automated calling and live calling scripts. Discuss your creative workflow