New Wesleyan Media Project tracking finds 220 AI-made or AI-enhanced political ads across 34 states this cycle, with researchers saying the true count is likely higher. Only about a quarter disclose the AI, and Republican-aligned sponsors account for most of the spending.
2026-10-16 · by Moe Ameen
Spending on political ads created at least partly with AI has climbed past $137 million ahead of the November 3, 2026 US midterm elections, according to tracking by the Wesleyan Media Project reported on October 16, 2026. The group has identified 220 such ads across 34 states so far, and its researchers caution the real figure is almost certainly higher because it is often difficult to prove AI was used. For scale: co-director Erika Franklin Fowler said the same project identified under $50,000 in AI-generated ad spending in the entire 2024 cycle.
The headline finding is the disclosure gap. Only about 27% of the identified ads carried a disclaimer telling viewers the ad featured AI-generated content; the large majority ran with no label at all. A separate CNN analysis earlier in the cycle reached a similar conclusion from the TV-ad side, finding that more than two-thirds of the money behind ads using AI-generated or manipulated imagery went to spots that did not disclose it. Spending also skews heavily partisan: Republican candidates and aligned groups accounted for roughly 81% of the identified ads and about 88% of the spending, with Democrats behind a much smaller share. (An earlier Wesleyan release in September put the running totals lower — around $80 million and roughly 164–170 ads — so treat the dollar figure as a moving tally that depends on the reporting date.)
The ads range from obvious parody to synthetic clips built to mislead. Coverage cited a fabricated video placing New York figures together in an invented advertisement, using their voices and likenesses with no AI disclaimer, alongside spots depicting events that never happened. Some campaigns have defended their ads as comedy or obvious satire; researchers have taken to calling the broader wave "slopaganda," cheap synthetic content meant to land an emotional impression rather than report a fact.
Regulation is a patchwork. Roughly 30 states have passed laws requiring AI disclosures in political ads, but there is no federal rule in force for the 2026 cycle, and requirements vary enough that an identical ad can be legal to run unlabeled in one state and not another. Exact totals differ by source and date, so confirm the current figures against the Wesleyan Media Project's own releases before citing a specific number.
This is a political-advertising story, but the operating lesson generalizes to every creator and brand using AI: the line that now separates trusted content from slop is whether it is consistent, clearly yours, and honestly disclosed. The failures in this report are undisclosed [deepfakes](/glossary/deepfake) of real people; the opposite — and the defensible way to use generative tools — is net-new content built on an identity you own and label as AI where the platform or the law asks. [Kompozy](/) is built for that side of the line. You generate content from your own AI Influencer persona, not a scraped likeness of someone else: a face-locked [Persona Short](/glossary/persona-shorts), a carousel, a quote graphic, a blog, all governed by a single [Persona Brief](/glossary/persona-brief) so a week of posts reads as one consistent brand instead of a scatter of synthetic one-offs. That consistency is exactly what "slopaganda" lacks and what an audience rewards.
Because the disclosure rules are a per-platform, per-state patchwork, distribution is where discipline pays off. [Autopilot](/glossary/autopilot) schedules and publishes across the eight social platforms plus blog and email from one queue, behind a per-post review gate where you confirm captions before anything ships — and for the platforms that require it, Settings → Publishing has a per-account AI-disclosure toggle (TikTok, YouTube) so connected accounts carry the platform's own AI label automatically instead of you retrofitting a disclaimer onto twelve copies of the same clip after the fact. For the labeling specifics advertisers are now navigating, see our guide on [AI-generated ad disclosure](/guides/ai-generated-ads-disclosure-and-ugc-creatives). The takeaway from the midterm data is not "AI ads work" — it is that undisclosed, impersonating content is becoming a liability, and the creators who win the trust war are the ones producing clearly-branded, clearly-disclosed content at the same speed.
Tracking by the Wesleyan Media Project, reported on October 16, 2026, put spending on political ads created at least partly with AI past $137 million ahead of the November 3, 2026 midterms, across 220 identified ads in 34 states. An earlier September release had lower running totals (around $80 million and roughly 164–170 ads), so the figure is a moving tally that depends on the reporting date, and researchers say the true number is likely higher.
It depends where the ad runs. Around 30 states have passed laws requiring AI disclosures in political ads, but there is no federal rule in force for the 2026 cycle, so requirements vary by state. In practice only about 27% of the AI ads the Wesleyan Media Project identified carried a disclaimer, meaning most ran with no label.
It is a term researchers have used for the wave of cheap, AI-generated political content in this cycle — synthetic clips and images meant to land an emotional impression or "deeper truth" rather than report a factual one. It spans obvious parody to misleading deepfakes, and the common thread in the criticism is that much of it runs without disclosing that AI was used.
It signals that AI-made content is now mainstream and actively scrutinized, and that platforms and regulators are pushing disclosure. The practical takeaway is to build content on an identity you own rather than impersonating anyone, keep it consistent, and add the AI label each platform or state expects. Transparency and brand consistency are becoming trust advantages, not costs.