How to Spot AI Fake News Before the November Midterms
PHILADELPHIA — With 29 days until the Nov. 3 midterm elections, voters are entering the stretch when fabricated political content spreads fastest. Artificial intelligence now lets influence campaigns manufacture convincing videos, images, and audio faster and more cheaply than old-fashioned propaganda, and social networks can push that material to millions of people within hours. Knowing how to spot the fakes has become basic civic self-defense.
Researchers cited by Agence France-Presse have linked an AI-driven influence operation to the 2026 midterms, saying it has circulated manipulated videos using cloned voices and unauthentic news branding. According to NewsGuard, which runs a tracker of false election claims, recent fabrications have included an AI-generated roadside billboard image and a doctored campaign photo. The campaigns work because they look real at first glance, and most people never look twice.
The first habit to build is checking the source before believing the claim. Ask who posted the material and whether that account is who it claims to be. A video attributed to a familiar news brand may carry that brand’s logo without any connection to the actual outlet, a trick researchers say has been used in this cycle. When something seems designed to provoke outrage, pause and look for the same story on the websites of established news organizations before deciding it is real.
The second habit is checking images the way fact-checkers do. NewsGuard analysts used reverse image searching to trace one viral billboard photo back to a satirical original posted months earlier, showing that many sharers had stripped away its context. Anyone can do a simpler version of the same check by running a suspicious image through a reverse image search and seeing where else it appears and when it first showed up online. If the earliest appearance is a joke page or the photo predates the election, treat the claim with deep skepticism.
Watch viral clips especially closely, and slow down on video that arrives through a friend’s feed with an explosive claim attached. NewsGuard’s analysis of one fabricated campaign image pointed to telltale signs: distorted faces in the background and a campaign pin that did not match the candidate’s real pins. AI-detection tools flagged that image with near certainty as machine-generated, according to NewsGuard. Look for odd details like warped text, mismatched logos, or faces that blur at the edges.
Be careful about asking AI chatbots for election answers. A 2026 study by the Institute for Strategic Dialogue tested five major chatbots on election questions and found accuracy ranging from 60 to 84 percent, with results another 16 percentage points worse in Spanish. An independent analysis cited in academic research concluded that chatbots gave a significantly inaccurate answer to midterm questions in roughly 90 percent of cases, often because of flawed source retrieval rather than deliberate deception. For voting facts like polling places and deadlines, go to an official county election office instead.
State laws are beginning to require disclosure, though their limits matter. A Louisiana law that took effect in June 2026 requires clear warnings on campaign ads that use AI, and a separate state rule requires AI-generated voice calls to disclose that use at the start of the call. But according to an analysis of those laws, the ad-disclosure rule does not cover candidates for federal office, and a disclosure label says how the content was made, not whether its claims are true. Rules vary by state and do not reach everything.
The simplest protection remains an old one. Do not share political content you have not checked, especially in the final month of a campaign when volume and emotion both peak. Cross-check any dramatic claim against at least one established news outlet, and treat unsourced viral posts as rumors until proven otherwise. In a season when machines can fake voices and faces, a little patience is the most reliable filter.