Deepfakes and Election Integrity: AI’s Most Concrete 2026 Worry

One-line job: Understand why AI misinformation is the concern where public and experts fully agree — and what’s actually been documented so far. Audience: Operators, voters, and anyone building content systems this election cycle. Not for: Readers wanting a technical deepfake-detection guide. This is the evidence and sentiment picture. Last verified: 2026-08-13 Evidence weight: documentation-verified

Most AI anxieties split along an expertise line — the public worries about one thing while the people building the technology worry about another. Misinformation is the exception. Pew Research (April 2025) found 66% of U.S. adults and 70% of AI experts worried about AI-driven inaccurate information. When the people who build the models worry at higher rates than everyone else, that’s not hype. That’s insider knowledge.

Why elections are the concrete case

Abstract concern becomes specific in election years. Jumio’s 2024 Online Identity Study (n=8,077 across the U.S., U.K., Singapore, and Mexico) found 72% concerned about deepfakes influencing elections.

And the interference isn’t hypothetical — it has a documented proof of concept predating this cycle. In February 2024, an audio deepfake of President Biden was used in an automated robocall urging New Hampshire primary voters not to vote. The incident produced the first wave of enforcement attention and demonstrated the full pipeline: cloned voice, cheap automation, targeted distribution, real confusion on election day.

What the surveys stack up to

The concern cluster is consistently top-tier:

• Pew (April 2025): 66% of adults / 70% of experts worried about AI inaccurate information
• Reuters/Ipsos (Aug 2025): misinformation among the most-cited concerns, and the fastest-rising year-over-year per Stanford HAI’s aggregation — though note the Reuters URLs sit behind an authwall; figures here are confirmed via Pew and Stanford aggregation rather than the primary release
• KPMG 2025 global study: manipulation or harmful use ranks as a top-tier risk across 47 countries

Why alignment matters

Here’s what makes this different from the other big AI concerns. On job displacement, the public-expert gap is 50 points — which means every public conversation starts with a credibility fight about whether the concern is even valid. On misinformation, there’s no such fight. Experts aren’t telling the public to calm down; they’re roughly as worried as everyone else.

That changes the operational calculus. Products touching political content, synthetic media, or user-generated video will face scrutiny not just from regulators but from users who already believe the risk is real — because the experts they’ve heard from said so too.

The call

1. If your product handles media or political content, assume provenance questions are coming. Content credentials and watermarking standards exist now; adopting them early is cheap.
2. Don’t wait for regulation to define your synthetic-media labeling. The 2024 NH robocall set the precedent that consequences arrive after the damage.
3. If you operate platforms, your moderation choices this cycle will be judged against the 72%-concerned number, not against what’s technically feasible.
4. Watch the expert signal: when insiders worry at rates matching outsiders, treat the concern as load-bearing rather than manufactured.