Society & Ethics

Campaigns Have Spent $16 Million on AI Ads Ahead of the 2026 Midterms

Campaigns and groups have spent over $16 million on AI political ads this year, a CNN analysis found, weeks before the Nov. 3 midterms put all 435 House seats on the ballot.

By Rebecca Stone5 min read

Updated

Why it matters

  • Campaigns and groups have spent more than $16 million on AI political ads this year, per a CNN analysis.
  • The 2026 midterm election is Tuesday, Nov. 3, with all 435 House seats and 33 Senate seats on the ballot.
  • The consultant behind AI Biden robocalls in New Hampshire's 2024 primary was fined $6 million for voter suppression.
  • Free detection tools include OpenAI's Verify, Google's SynthID Detector and the Content Authenticity Initiative's Inspect tool.
  • Pangram charges individual users $20 per month for AI content identification.

Political campaigns and outside groups have spent more than $16 million on AI-generated or AI-edited advertising this year, according to a CNN analysis, as the 2026 midterm election approaches on Tuesday, Nov. 3.

The figure lands in the final weeks of a campaign cycle in which AI-manipulated media has already surfaced in high-profile races. One ad gave Georgia Senator Jon Ossoff AI-applied cosmetic procedures. Another fabricated Texas Senate candidate James Talarico appearing to deface a Bible. Some of these ads carried disclosures about their AI use. Others did not.

The stakes are structural, not cosmetic. All 435 seats in the House of Representatives are on the ballot, along with 33 Senate seats and thousands of state and local races. Voters navigating that volume of content now face media that, according to CNET's reporting on the trend, is "nearly impossible to distinguish real from AI" after two years of rapid model improvements.

What happened in 2024 — and why it matters now

The 2026 cycle is not the first collision between generative AI and American elections. During the 2024 presidential primary, thousands of New Hampshire voters received robocalls using an AI voice replica of Joe Biden urging them not to vote. The Biden campaign had not sanctioned the calls. AI robocalls are illegal. Authorities charged the Democratic political consultant behind the operation with voter suppression and fined him $6 million.

That case established the enforcement baseline. The technology has improved since. As CNET puts it, improvements to AI since 2024 "only make for sharper swords in a bad actor's arsenal."

Can voters still detect AI content?

The visual tells that once exposed AI images — seven-fingered hands, warped text, mismatched ears — have grown scarce. Model quality has closed those gaps. The result: eyeballing a photo or video is no longer a reliable verification method.

Detection now runs through metadata, not appearances. Several free tools scan image, video and audio files for hidden signatures that indicate AI generation:

  • The Content Authenticity Initiative's Inspect tool identifies multiple kinds of AI content.
  • OpenAI's Verify tool detects content made with ChatGPT.
  • Google's SynthID Detector identifies Google-made AI content; users can also upload files to Gemini to scan for SynthID watermarks.

None of these is definitive. CNET warns that any AI detection service claiming a 100% success rate is lying, citing too many AI services and too little industry standardization for any tool to succeed all the time.

Paid options exist. Pangram offers AI content identification to individual users at $20 per month. Apple is adding a verification feature called Reference Images, though it is limited to the newest iPhone 18 Pro models.

The detection ecosystem also remains fragmented by vendor. SynthID Detector catches Google-made content. OpenAI's Verify tool catches ChatGPT output. A deepfake built on an uncatalogued open-source model may slip past all of the free checkers at once.

What should voters do when they receive election messages?

Cybersecurity research going back decades shows the same pattern: attackers prey on emotion to cloud judgment. The classic election-season version arrives as a text message claiming a polling location has changed, with a link to the "new" site. CNET's guidance is blunt — do not click. Instead, verify the claim through official channels.

The verification path runs through several institutions:

  • Your state's board of elections website carries the most official and current information on polling locations and hours.
  • The US Election Assistance Commission publishes a map with each state's official election contact information.
  • Governor's office websites often post voting information.
  • The nonprofit Vote.org provides ballot information for voters in any state and offers a free tool to check registration and preview ballots.

For candidate positions, CNET recommends starting with campaign websites and published platforms rather than images or videos circulating online. The advice carries a caveat the source states plainly: there is no guarantee a candidate posts entirely truthful information, but an official platform at least reflects the campaign's stated positions on major issues.

Why the timing matters

Early voting periods begin in mid-October in many states, which compresses the window for voters to prepare. CNET's practical checklist ahead of that deadline:

  • Make a voting plan now rather than in the final days.
  • Check your state's voter ID laws in advance.
  • View your ballot before election day and research every candidate on it, not only the headline races.
  • Use fact-checking resources such as Snopes.com when something looks wrong.

Voters who encounter problems at their polling place can call the nonpartisan Election Protection Hotline at 866-687-8683.

What does the $16 million figure signal?

The spending number matters because it measures adoption, not experimentation. Campaigns do not put eight figures into a tactic that does not work. The ads targeting Ossoff and Talarico show the two dominant use cases already in the field: cosmetic mockery to make an opponent look absurd, and fabricated actions to make an opponent look scandalous.

Enforcement, by contrast, remains reactive. The New Hampshire robocall case produced a $6 million fine — but only after the calls reached voters. Disclosure rules for AI-generated campaign ads vary, and CNET's reporting notes that some 2026 ads ran with no AI disclosure at all.

The gap between generation and detection defines the 2026 cycle. Free verification tools from OpenAI, Google and the Content Authenticity Initiative now exist, but each covers its own ecosystem's output. Pangram's paid tier and Apple's Reference Images on the iPhone 18 Pro extend coverage only partly. The realistic defense, per CNET's guidance, is procedural: verify against official sources, plan the vote early, and treat any single image, video or call as unproven until checked.

With weeks to go before Nov. 3, that burden now sits with 435 House districts' worth of voters.

Original: vote.org

Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Correspondent covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.

289 articles

Related articles

  1. Campaign Finance Filings Expose What Politicians Spend on AI
  2. OpenAI Says It Blocked 250,000 Election Deepfake Requests
  3. AI Policy Becomes Early Flashpoint for 2028 Democratic Field
  4. Princeton Researcher Used a $20 AI Subscription to Break Georgia Ballot Secrecy
  5. Trump Renames AI "Super" and Brands Resisters "The Enemy"

« Previous article