AI fakes in the 2026 midterms: the tracker

Reviewed by Colin Balfe on Oct 27, 2026 · Updated Oct 27, 2026 · Election Day is Tuesday, Nov 3 · Every entry sourced
The short answer

13 AI-made election fakes and AI ads are on our record this cycle, 9 of them confirmed fake or taken down. Before you share a shocking clip of a candidate this week, run the checks below. For where and when to vote, use your state election office or vote.gov, not a video, text or call.

Before you share: how to check a clip

  1. Pause before you shareIf a clip makes you angry or shocked, slow down. That is exactly when fakes spread fastest. (source)
  2. Find where it first came fromAsk who posted it first and where. A brand-new or anonymous account is a warning sign; the account behind the Talarico deepfake was created only a month earlier. (source)
  3. Reverse-search a screenshotTake a screenshot of a key frame and upload it to Google Images or TinEye to find the original and see whether it was edited. (source)
  4. See what trusted outlets sayLeave the post and search for the claim. If something this big were real, several news outlets or fact-checkers would be reporting it, or the campaign would have confirmed it. (source)
  5. Look and listen for glitchesCheck whether lips match the words, and watch for odd blinking, warped edges or backgrounds, and a flat or slurred voice. Newer fakes may show none of these signs, so this is only one clue. (source)
  6. Check for Content CredentialsUpload an image or paste its link into contentcredentials.org/verify to see any record of the tool used, including AI. A missing credential does not prove a file is fake. (source)
  7. Read the fine print on adsCheck who paid for an ad (usually stated at the end) and look for an AI disclaimer. Many states require one, but most AI ads this year carried none. (source)
  8. Get voting details only from officialsDon't click links in unexpected texts or emails about your registration. Check with your state or county election office instead. (source)

The fakes so far

Newest first. Confirmed fake: a fact-checker, platform, official or the poster confirmed AI. Labelled AI by poster: an ad that says it used AI. Alleged: claimed, not confirmed.

  • removed Fort Bend County Judge. An AI-generated video of four Black residents, one of them a Wong poll worker, praising interim county judge Daniel Wong for giving them jobs was shared on Facebook. Made or posted by: Shared by Wong's campaign page and the Fort Bend Republican Party; the campaign says, without naming anyone, that 'a Democrat' made it. How it was caught: Commenters noticed warped backgrounds and clothing and the same-sounding voice for all four speakers; the campaign then took it down. (Chron via Yahoo News, Oct 25) The case file →
  • confirmed fake Texas US Senate. Two videos posted on X show a man posing as a former student of James Talarico and falsely accusing him of sexually assaulting him in 2012; researchers judged the videos 'very likely' deepfakes. Made or posted by: Russia-aligned influence operation Storm-1516, per Graphika ('high confidence') and Clemson's Darren Linvill; posted by the new X account @zyon_lewis_5, which impersonated a real former student. How it was caught: Graphika and NewsGuard flagged lips out of sync with the audio, stiff faces and no body movement, plus AI-detection results; the real man told Lead Stories it was not him; PolitiFact rated it Pants on Fire; X suspended the account by Oct 22. (PolitiFact, Oct 22) The case file →
  • confirmed fake Texas US Senate. A campaign ad uses AI to depict James Talarico defacing a Bible with crayon, with the line 'In the beginning God was non-binary.' Made or posted by: Ken Paxton's (R) Senate campaign. How it was caught: Identified in an Axios review of AI political ads. Talarico said he had tried to ban AI deepfakes in campaigns 'because they are misleading'. (Axios, Oct 19)
  • labelled AI by poster Michigan US Senate. Super PAC ads show a fabricated family photo of Abdul El-Sayed with podcaster Hasan Piker and four women in niqabs, and an AI scene of him playing a Risk-style game with cards reading 'Death to America' and 'Celebrate 9/11'. Made or posted by: Front Line Action (super PAC). How it was caught: The family-photo ad carries a written and spoken AI disclosure. The group says the ads are 'grounded in El-Sayed's own public statements'. (Axios, Oct 19)
  • labelled AI by poster Pennsylvania Governor. An ad shows an AI version of Gov. Josh Shapiro talking about taxes, abortion and data centers and 'admitting' he fast-tracked data centers after taking $5 million in contributions. Made or posted by: Stacy Garrity's (R) campaign. How it was caught: The ad itself says it was made with AI. Shapiro's spokesperson called it 'cheap, dishonest AI' (expletive). (Axios, Oct 19)
  • labelled AI by poster New York Governor. A series of campaign ads uses AI versions of Gov. Kathy Hochul, including one billed as 'what a truthful Kathy Hochul ad would look like', in which a man outside a police precinct asks her 'Want some crack?', and earlier ads showing her gardening with NYC Mayor Zohran Mamdani. Made or posted by: Bruce Blakeman's (R) campaign. How it was caught: Some of the ads carry AI notices, but how clearly varies. In August the state Democratic Party filed a complaint with the state Board of Elections alleging that some lacked the labels state law requires; Blakeman calls them satire. (Axios, Oct 19)
  • confirmed fake Kentucky 6th District (US House). Several AI deepfakes of Democrat Zach Dembo, including one of him staring into the camera, ran in ads boosting Republican Ralph Alvarado. Made or posted by: No Going Back PAC, backed by MAGA Inc.. How it was caught: Identified in an Axios review. Dembo: 'I did not expect them to lie about my appearance'. (Axios, Oct 19)
  • confirmed fake Congressional races (Democratic candidates). AI-made video clips showed actors and musicians, among them Robert De Niro, Julia Roberts and Cher, attacking Democrats; some pushed the hashtag #AllDemocratsAreCriminals, and many other clips made false corruption claims against Democratic candidates. Made or posted by: Russian state-backed operation Matryoshka (also called Operation Overload), per social media researchers cited by Bloomberg. How it was caught: Social media researchers cited by Bloomberg. The clips began appearing on X in early September. (Bloomberg, Oct 16) The case file →
  • labelled AI by poster Colorado Proposition NN (ballot measure). An AI-generated video shows state Sen. Cathy Kipp as a grocery cashier who pockets customers' change 'to help educate children', attacking the ballot measure she sponsored. Made or posted by: Independence Institute (libertarian-leaning think tank). How it was caught: The video admits near the end that it is AI but lacks the formal disclaimer. Kipp filed the first complaint under Colorado's deepfake disclosure law; the Secretary of State's elections division is investigating, and the Institute calls it satire. (Colorado Politics, Oct 13)
  • removed Mississippi US Senate. An AI-generated video showed Sen. Cindy Hyde-Smith running away from reporters who were asking about her healthcare vote; it was captioned 'Based on a true story.' Made or posted by: Scott Colom's (D) Senate campaign, on Facebook and Instagram. How it was caught: The state Republican Party complained. AG Lynn Fitch said the video 'appears to violate' Mississippi's political deepfake law and set a removal deadline. Colom called it parody but took it down on Oct 23. (Commercial Dispatch, Oct 23) The case file →
  • confirmed fake Wisconsin US House (Rep. Derrick Van Orden's seat). An AI-generated video, posted on X in August, shows Democratic challenger Rebecca Cooke saying she wants to 'take advantage of as much corruption as possible and give every illegal alien free healthcare', which she never said. Made or posted by: Rep. Derrick Van Orden (R), via his X account. How it was caught: Cooke's lawyers sent a cease-and-desist letter. Van Orden replied with an AI video of a duck saying 'cease and desist'. No removal is reported. (Axios, Oct 2)
  • removed Texas Lieutenant Governor. Two attack ads used AI-generated images of state Rep. Vikki Goodwin at a University of Texas football game wearing a shirt reading 'taxes' instead of 'Texas'. Made or posted by: Lt. Gov. Dan Patrick's (R) campaign; a spokesperson called them 'obvious parody produced to entertain'. How it was caught: Goodwin said she never did or said what the ads show and filed a criminal complaint under Texas's election deepfake law with the Travis County Sheriff on Sept 19; Patrick deleted both videos. (FOX 4, Sep 22)
  • confirmed fake Georgia US Senate. A series of ads uses AI renderings of Sen. Jon Ossoff getting hair and beauty treatments in a salon, with the line 'Jon Ossoff talks pretty, his actions ugly.' Made or posted by: Senate Leadership Fund (super PAC aligned with Senate Republicans). How it was caught: Reporters identified the AI renderings. The ads carry no AI label, and Georgia law does not require one. Ossoff called them 'deepfake videos used to fabricate political advertising' and said they 'should come down immediately'. (Axios, Oct 19)

The numbers

$137 million on 220 AI adsThe Wesleyan Media Project counted at least 220 political ads made or enhanced with AI by Oct 11, costing just under $137 million, against less than $50,000 of AI ad spending in the 2024 presidential election.NBC News via Yahoo, Oct 23
Only 27% disclosed AIOf those 220 ads, only 27% told viewers AI was used. 81% were paid for by Republican candidates or groups and 12% by pro-Democratic sponsors.NBC News via Yahoo, Oct 23
More than 4 in 10 voters say an AI ad fooled themIn a Scripps News/Talker Research poll of 1,000 registered voters (Sept 30–Oct 1), more than four in ten said they had seen something in a political ad they thought was real and later learned was AI. Another 30% weren't sure. The headline says 43%.Scripps News, Oct 14
77% of Democrats and 79% of Republicans want rulesThe same Scripps poll found more than three-quarters of voters in both parties want rules on the use of AI in politics.Scripps News, Oct 14
31 states have election-deepfake laws, but no federal lawNCSL's June 2026 count: 31 states regulate election deepfakes. 28 require a disclosure, and Maryland, Minnesota and Texas ban certain deepfakes even when labelled. Congress has set no federal standard.The Conversation via Fast Company, Oct 3
FBI: scammers use AI to impersonate candidatesAn FBI public service announcement (Oct 15, 2026) warns that scammers use AI to generate 'highly realistic voices, images, and written messages' impersonating candidates, campaigns and government bodies, and push fake PACs, 'campaign investment pools' and texts falsely saying you are not registered.FBI IC3, Oct 15

The rules

AI-voice robocalls fall under federal robocall law
The FCC ruled (adopted Feb 2, released Feb 8, 2024) that AI-generated voices count as 'artificial' voices under the Telephone Consumer Protection Act, so such calls need the called person's prior express consent. FCC Chair Rosenworcel said state attorneys general 'can go after the bad actors behind these robocalls'. (FCC Declaratory Ruling FCC 24-17)
Texas: no deceptive AI ads in the final 30 days
Texas's 2019 law bars AI deepfake political videos published within 30 days of an election with intent to influence it. Rep. Vikki Goodwin filed a criminal complaint under it against Lt. Gov. Dan Patrick's campaign in September. (FOX 4, Sep 22)
Mississippi: criminal penalties for election deepfakes
Section 97-13-47 of the Mississippi Code carries up to 5 years in prison and/or fines up to $10,000 when a deepfake is meant to deter someone from voting. The attorney general used it to force down a Senate campaign's AI video on Oct 23. (WLBT, Oct 23)
Colorado: label it or face civil penalties
Colorado's HB 24-1147 requires a clear disclosure on deepfakes of candidates, with exemptions for satire or parody. Its first complaint, over an AI video of state Sen. Cathy Kipp, is under investigation. (Colorado Politics, Oct 13)
Maryland: a label doesn't make it legal
A Maryland law in effect since June 1, 2026 bans certain deceptive election deepfakes even if labelled, with fines up to $5,000 and up to five years in prison. In September a federal judge temporarily blocked Montana's deepfake law on First Amendment grounds. (The Conversation via Cobb Courier, Sep 30)

Voting: the only sources to trust

  • Election Day is Tuesday, Nov 3, 2026 The midterm elections are held on November 3, 2026. (Al Jazeera, Oct 9)
  • Check your registration at vote.gov vote.gov is an official US government website for registering, checking your registration or updating it in your state. The FBI's Oct 15 alert recommends confirming your status there rather than through a link in a text or email. (FBI IC3, Oct 15)
  • Find your own election official at CanIVote.org The National Association of Secretaries of State's #TrustedInfo2026 campaign points voters to their state and local election officials, reachable through its nonpartisan site CanIVote.org. (NASS #TrustedInfo2026)
  • Verify any call or text saying your polling place moved Don't trust a call, text or email about where or how to vote without checking it with your local election office. In a 2020 example, Oklahoma's state election board secretary said officials 'will never contact you by text message to tell you your polling place has changed'. Report suspected voter intimidation to 866-OUR-VOTE. (Consumer Reports, Oct 28, 2020)

Questions people ask

How can I tell if a political video is a deepfake?

Don’t rely on spotting glitches. Check where it first appeared, whether a major outlet or the campaign has confirmed it, and whether fact-checkers have rated it. A brand-new or anonymous account posting an explosive clip is a warning sign.

Are AI political ads legal?

In most states, yes, with conditions. Many states now require AI-generated political ads to carry a disclosure, and a few ban certain deceptive deepfakes close to an election. There is no federal law covering AI in campaign ads; AI-voiced robocalls are covered by federal robocall rules.

What do I do if I see an election deepfake?

Don’t share it. Report it on the platform where you saw it, and if it gives false voting information, report it to your state or local election office.

Where can I check where and when to vote?

Your state or local election office, or vote.gov. Treat any call, text or video that changes your polling place or date with suspicion and check it with your election office.

Related: Is this video real? How to check before you share

Sources

Spot an error or something out of date? Tell us; corrections are made and logged. Written with AI assistance from public sources; every fact checked and approved by Colin Balfe. General information, not legal, medical or financial advice.