Understand state and federal laws on AI political ads in 2026, and learn how to check a video's provenance when the disclosure label is missing or hidden.
This article was researched and drafted with AI, then checked and released by a person before it went live. The illustration is AI-generated. Every factual claim links to its source, so you can verify it yourself.
AI-generated illustration. A television screen showing a political debate with a magnifying glass graphic overlaid on a paused video frame.
An AI generated political ad disclosure is required by law in 30 states as of 2026, but the rules vary wildly from year-round bans to tiny on-screen text. Because federal regulation remains stalled, voters must often verify an ad's provenance themselves using metadata and earliest documented publication records.
When a sensational video of a political candidate floods your timeline right before an election, you might expect the law to require clear labeling. Searching the screen for an ai generated political ad disclosure has become a standard reflex for voters. The reality of election law in 2026 is far more complicated than a simple federal mandate. Depending on where you live, that video might be strictly illegal, perfectly legal, or operating in a grey area where tiny, unreadable text technically satisfies a state requirement.
Understanding how these rules work is the first step to navigating campaign season. When the legal disclosures fail or campaigns simply ignore them, you need to know how to check the provenance of a video yourself.
There is currently no binding federal law that specifically regulates artificial intelligence in political advertising across the United States. Regulatory agencies have debated the issue, but broad federal rules have not materialized for the 2026 midterms.
The Federal Election Commission voted in September 2024 not to open a formal rulemaking process on the use of AI in campaign ads. Instead, the agency adopted an Interpretive Rule clarifying that existing regulations on fraudulent misrepresentation already apply to AI. This means a candidate cannot use AI to falsely speak or act on behalf of another candidate in a damaging way, but it does not mandate a universal disclosure label on all synthetic media.
Other agencies have attempted to step in with limited scope. The Federal Communications Commission proposed disclosure rules for radio and television advertisements in August 2024. However, these rules do not cover the internet, which is where most viral deepfakes circulate. The FCC has also maintained strict boundaries on automated calls, confirming in October 2026 that it has zero plans to ease political robocall rules to allow AI-generated voice campaigns to reach consumers unchecked.
Because federal action has stalled, the rules governing synthetic media depend entirely on state legislatures. As AuditSocials reported, 30 states had enacted statutory regulation of AI or synthetic media in political advertising by May 2026. This state patchwork is the binding framework for the midterm election cycle.
These state laws are not uniform. Axios noted in August 2026 that voters face highly uneven protections. While laws in California and Hawaii were permanently enjoined, 29 other states have active rules in effect. The approaches generally fall into a few distinct categories:
Michigan provides a concrete example of how specific these disclosure rules can be. Its law requires a three-second spoken disclaimer plus a four-second text overlay in the 90-day window before an election. Yet, despite these highly specific statutes, AP News found that candidates continue to use AI ads without disclosures entirely.
Even when campaigns comply with state disclosure laws, the resulting labels are often criticized as inadequate. A prime example occurred in the Texas Senate race. The National Republican Senatorial Committee released an ad in March 2026 targeting Democratic candidate James Talarico.
The video used artificial intelligence software to visualize Talarico seemingly reading his own past social media posts. As CNN reported, the ad featured a phony version of the candidate talking directly into the camera. The campaign did include an "AI GENERATED" label on the screen to comply with disclosure expectations. However, experts speaking to UC Berkeley pointed out that the text was easy to miss, raising serious questions about whether voters could still be misled despite technical compliance.
When the law fails to provide a clear label, voters must verify the media themselves. This is where provenance tools come in. The wyper AI Spotter is built specifically to provide a provenance report for images and videos. It never outputs a percentage of how "AI-ish" something looks. A percentage is a guess, and guessing is the exact harm we exist to prevent.
Instead, the tool checks concrete layers: content credentials (C2PA) verified against 55 official trust anchors, file metadata, reverse image search, earliest documented publication, and quote verification. For video, the AI Spotter takes still frames and searches each one separately. A hit on any single frame is a hit on the video.
It is vital to understand how content credentials behave on social media. A C2PA signature sits in its own box beside the picture data. Any repackaging of the file drops the signature, including a lossless repackaging that leaves every pixel untouched. Because social media platforms automatically repackage files upon upload, a video taken off a feed will almost always return nothing on the credential layer. That is the normal case, not a suspicion of tampering. The report states this clearly.
What actually carries the weight for video verification is the earliest appearance record and the gap between the event date and the post date. This method catches the most common form of real video deception: old, authentic footage relabeled as a new event. A pixel detector cannot see that deception at all.
Please note that the AI Spotter requires heavy processing and is not in the free tier. It costs 1 credit per image and 4 credits per video on paid plans, because a provenance report calls three paid services on every single run. However, checking the text claims made inside a video remains entirely free. You can run 5 text checks a day with no account required on the free wyper web app or via the browser extension.
| Manual checks | wyper web app | wyper extension | |
|---|---|---|---|
| Cost | Free | Text check is free, AI Spotter is paid | Text check is free, AI Spotter is paid |
| Data privacy | Data never leaves your browser | Requires sending the URL or text | Requires sending the URL or text |
| Video AI Spotter | Requires manual reverse search | Yes (paid credits required) | Yes (paid credits required) |
| Text fact-checking | Requires manual search queries | Yes (Dual-AI cross-check) | Yes (Dual-AI cross-check) |
| Subtitle reach | Limited to visible text | Limited by lack of session access | High (uses your logged-in session) |
Who needs which: Voters who want to verify a specific claim without installing anything should use the web app for a quick text check. Researchers and journalists who frequently encounter suspected deepfakes across different platforms will benefit from the extension, which can reach subtitles behind logged-in sessions and offers seamless access to the paid AI Spotter for video provenance. Users who prioritize absolute privacy and have the time to perform their own reverse image searches should stick to manual checks.
If a political ad features a candidate speaking, you can check the transcript of what they are saying using wyper's text fact-checking tools. Both products return a truth score of 1-10 alongside an evidence chain of real, linked sources.
The standard check runs a Dual-AI cross-check using Gemini and Grok. Grok searches the web for itself during this process, ensuring the two models do not share a single evidence set. When we measured 30 claims using frozen sources, Gemini and Grok agreed in 29 out of 29 cases. Because two models trained on overlapping internet text are correlated rather than truly independent, confirmation by two AIs carries little unique information.
If you require deeper scrutiny, the Full Spectrum option provides a Triple-AI reading: three models, one evidence set. The third voice does not search. It receives exactly the same sources the main model received and reads them independently. Therefore, a disagreement is a different reading of the same material, not a different search result. In our testing, a third model of different origin disagreed in 3 out of 26 cases, every time citing a machine-verified quote.
Our system prioritizes primary sources. In a test of 24 claims with the root named in advance, the tool reached primary sources 63 percent of the time and legacy media 14 percent of the time. When re-fetching 633 real citations, 73 percent were found verbatim on the page, though 19 percent of pages were unreachable for automated readers due to paywalls or PDF formats. Sources are never filtered by origin, meaning there is no mainstream bonus. The reader decides what to trust.
State laws currently dictate whether an AI generated political ad disclosure is required, resulting in a confusing patchwork of rules for the 2026 midterms. When campaigns use tiny labels or ignore the rules entirely, voters must rely on provenance tools and text fact-checking to verify the claims themselves. Always check the earliest documented publication date of a video to ensure old footage is not being used to manufacture a new political scandal.
The wyper Fact-Check extension runs these checks on the post itself: truth score, evidence chain, and the date gap that catches recycled footage.
The web app installs to your home screen in one tap. No store, no account.
Already using it? ★★★★★ Rate it on the Chrome Web Store. Reviews decide what others get shown first.
wyper