AI election deepfakes have moved from warning to everyday reality in the 2026 midterms. Campaigns have aired AI-made ads, chatbots answer voting questions for millions of people, and 31 states now regulate AI in campaigns. With Election Day on Tuesday, November 3, here is what the research says and how to check what you see.
Key facts (as of November 2026)
- The Wesleyan Media Project counted at least 164 AI-generated or AI-enhanced political ads with about $80 million in spending, SciLine reported.
- Most of those ads (69%) did not disclose that AI was used, according to the same count.
- 31 states have laws on AI in campaigns, and there is no federal rule as of October 2026, SciLine noted.
How are AI election deepfakes showing up in 2026?
Mostly in paid ads and on social media, and often openly rather than secretly. Al Jazeera described several examples this cycle. In New York, Republican governor candidate Bruce Blakeman ran ads showing Governor Kathy Hochul and New York City Mayor Zohran Mamdani in scenes that never happened. He says they are labeled satire. The state Democratic Party filed a complaint with the board of elections arguing they break disclosure rules.
Democrats have used the tools too. In San Francisco, state Senator Scott Wiener launched an AI chatbot that answered questions about his House race opponent, Connie Chan, then pulled it after criticism. He said it “missed the mark.” According to the same report, more than 80% of the spending on AI-based ads tracked by the Wesleyan Media Project came from Republicans.
Beyond campaigns, foreign influence networks are also in play. In an opinion piece for The Fulcrum, Lawrence Norden of the Brennan Center and David Evan Harris of UC Berkeley pointed to reports of Russian and Chinese networks using AI-made videos and automated accounts aimed at US voters.
Do AI election deepfakes actually change votes?
The honest answer is that nobody has measured it directly yet. A guide for journalists published by SciLine on October 26 summed up the evidence, and it is more nuanced than the headlines.
- Voters dislike AI deception, but it may not cost candidates. A February 2026 study in Political Communication found Americans strongly disapprove of it. However, that disapproval mostly showed up as support for regulation, not as lower ratings for the candidate.
- Chatbots can be persuasive. A December 2025 study in Nature found a chatbot conversation shifted voters’ views of candidates by a few points on a 100-point scale. That is larger than typical TV ad effects.
- More persuasive often means less accurate. A large UK study in Science found the most persuasive AI models tended to give less accurate information.
There is also a quieter risk. Researchers call it the “liar’s dividend”: once people know fakes exist, a politician can dismiss a real recording as AI. In short, deepfakes can damage trust in genuine evidence as well as spread false claims.
Can you trust a chatbot for voting information?
Not on its own. About half of US adults now use chatbots to look things up, according to Pew data cited by SciLine. Yet a Brennan Center review in August found roughly half of chatbot answers to election-misinformation questions were inaccurate.
Safeguards also vary by company. When Norden and Harris tested four major chatbots, three refused to write prompts for realistic fake election images, while one produced them. So the rule is simple: use AI to understand an issue, then confirm the facts with an official source. The companies behind these tools also face new financial scrutiny, as our guide to Anthropic’s IPO filing shows.
What rules apply to AI in elections?
Rules exist mainly at the state level. Most of the 31 state laws require AI-made political ads to carry a label, though the details differ. There is still no federal law on AI deepfake ads, and Ilana Beller of Public Citizen told Al Jazeera that one is still needed.
Labeling is also becoming technical. California’s AI Transparency Act required AI companies to embed provenance data starting August 2, 2026, Norden and Harris noted. In Europe, Article 50 of the EU AI Act introduces similar marking duties for AI-generated content from the same date. OpenAI’s newest image tool, for example, adds both C2PA data and an invisible watermark, as our guide to ChatGPT Images 2.6 explains. That matters for readers following why the US midterms matter to Europe.
How can you spot AI election deepfakes before you vote?
- Check where it came from. Is the account official? If a clip shows a candidate saying something shocking, look for it on the candidate’s own channels or in reports from several major outlets.
- Look for the label. Many states require AI disclosures, often as a small line at the start or end of an ad.
- Ask why it is going viral. Content built to make you angry deserves a second look before you share it.
- Go to the source for voting rules. Your local election office is the authority on when, where and how to vote, as experts quoted by SciLine advise.
- Watch results from trusted broadcasters. On election night, rely on live coverage from established newsrooms rather than viral clips.
How to watch election night live
The safest way to follow results is live coverage from news organizations with their own decision desks. Our guide to watching the midterm results live covers every major network, and the C-SPAN live stream offers coverage without commentary. You can also track the odds on our midterm election forecast page and read about record campaign ad spending this year. AI also mattered at the ballot box in another way: our look at how data centers shaped the vote covers the races where the AI build-out was a campaign issue.
Coming soon: the NewsLive streaming app. We’re building one subscription for live international news on Android, Roku, Fire TV and desktop. The service is under development. See what the NewsLive app will offer and join the early-access list.
Frequently asked questions
Are AI election deepfakes illegal?
It depends on the state. 31 states regulate AI in campaigns, mostly through labeling rules, but there is no federal ban as of October 2026.
Which party uses AI ads more?
More than 80% of the money spent on the AI-based ads the Wesleyan Media Project tracked came from Republicans, Al Jazeera reported. Democrats have also used AI tools, as the Wiener chatbot shows.
Where can I check if a video is AI-generated?
Start with the original source and major news reports. Georgetown’s True Media project also accepts suspected AI media for analysis, though it requires sign-up.
