2024 was supposed to be the year of AI in elections. Reporters described it as "the first AI election," and the year opened with a deepfake of President Biden's voice robocalling New Hampshire primary voters.
I've spent a significant part of the last two years flying around the country to talk with election officials about AI, leading trainings and workshops and supporting the AI + Elections Clinic at Arizona State University. AI still comes up constantly in that work: at trainings, at conferences, in questions about what tools are safe to use and what threats to plan for. So with the 2026 midterms two weeks away, I expected AI to show up in the election news about as often as it did in 2024, maybe more. I wanted to check that against the numbers. It turns out AI is in the conversation, but it isn't in the headlines.
How I counted
The method is simple. Every weekday, electionline publishes a roundup of election news from across the country: national stories first, then stories sorted by state. It's one of the best single records of what election officials and the people who follow them are reading. If you don't already get it, sign up for electionline's daily emails. It's the easiest way to keep up with election news nationwide.
Thank you, electionline
None of this would be possible without electionline and Mindy Moretti, who is up long before most of us to read and sort the country's election news every single weekday. It's an extraordinary public service, and it's the reason a project like this can exist at all.
I collected the headline of every story electionline linked from January 2023 through October 20, 2026. That's just over 48,000 stories. Then I kept the ones whose headlines mention AI or a closely related term: "artificial intelligence," "deepfake," "chatbot," "generative," "OpenAI" and a few others. I read the matches by hand and removed ones that weren't really about AI, like ordinary robocall schemes and a candidate's claims about "mathematical algorithms." That left 201 stories.
There are real limits here. This counts AI headlines in electionline's roundup, not all reporting on AI and elections. electionline links a lot, but not every story, and it doesn't cover the broader national and tech press. I also only looked at headlines, so a story about AI under a headline like "New law targets election misinformation" doesn't count. A full-text review of every article would likely add more stories. And some of the calls about what counts as an AI story were judgment calls that another reviewer could make differently.
The goal is to show a trend, using a consistent way to measure how often AI shows up in one widely read source over time. The count can't tell us how much AI is actually being used in elections, or how often it causes problems.
What the headlines show
As a share of everything electionline linked, AI stories were about 2 in every 1,000 in 2023, about 9 in every 1,000 in 2024, and back to about 2 in every 1,000 in 2025 and 2026.
The fairest comparison is 2024 against 2026 over the same stretch of the calendar, since both are federal election years. From January 1 through October 20, electionline linked almost exactly the same number of stories in both years: 12,083 in 2024 and 12,223 in 2026. In 2024, 131 of them were about AI. In 2026, 25 were. Total election coverage held steady while the number of AI stories fell by about four-fifths.
We're not at Election Day yet, and the final weeks of a cycle bring a lot of election coverage. But that likely won't close the gap. In 2024, only 8 more AI stories ran between October 21 and the end of the year. And the gap was already wide this fall: from September 1 through October 20, there were 19 AI stories in 2024 and 8 in 2026.
The 2024 stories weren't one big burst, either. They ran all year: state officials warning about deepfakes in January, the New Hampshire robocall, tech companies pledging action in February, and a steady drumbeat of trainings, tabletop exercises and new state laws in between.
Why 2024 still stands out
Part of the answer is probably that 2024 was the first big test. Generative AI tools had only recently become widely available, the New Hampshire robocall showed what they could do, and election officials, lawmakers and reporters were all trying to get ahead of risks nobody had seen in a presidential election before. A new kind of threat makes news.
But I don't think the drop means people stopped talking about AI and elections. The bigger conversation is still very much about AI. Election officials are asking about it, training on it and writing policies for it. It just isn't showing up in this data. A few possible reasons, none of which a headline count can settle:
- AI use is becoming normal, and normal is less newsworthy. In 2024, an election office or candidate trying an AI tool was a story on its own. In 2026, the same thing is less likely to make the news, especially when it builds on something an office or campaign already tried in an earlier year. The neutral and positive uses of AI may be the ones fading from headlines fastest, even as they spread.
- The warnings are familiar now. The AI risks that dominated 2024 coverage have been widely reported. A story about AI risk is less likely to run unless something new happens.
- AI may be part of bigger stories. It can show up inside stories about cybersecurity, misinformation or office operations without making the headline.
- Much of the conversation happens where headlines don't: in trainings, peer networks, vendor pitches and internal policy discussions.
- The mix of stories in the roundup may have shifted. A daily roundup can only include so much, and what gets linked can change from year to year.
What the data can say is that AI showed up much less often in election administration headlines in 2026. It can't say AI matters less.
AI is likely touching elections more, not less
Other evidence points the other way: AI is reaching further into elections and election offices than it did in 2024.
- More election officials are using it. In the Brennan Center's 2026 survey of local election officials, 16 percent reported using AI in their work, up from 5 percent in 2024. Most use it for routine tasks like drafting social media posts and creating graphics. The share who want guidance from government agencies on using AI rose from 31 percent to 46 percent over the same period.
- More voters are using it. About half of U.S. adults now report using AI chatbots, according to Pew Research Center, up substantially from the summer of 2024. Reporters are finding that voters are turning to AI for election help, and the answers aren't always right.
- Election offices are putting it in front of voters. Ohio debuted an AI election assistant in late 2025. Arizona's secretary of state expanded its AI chatbot ahead of the midterms. Montgomery County, Pennsylvania, launched a voter services chatbot this month.
- And it's already causing problems. Election officials in Maryland pulled AI-generated posts. ProPublica reported that the FBI has looked at using questionable AI tech to review signatures on seized mail ballots.
In 2024, many of the AI headlines were about what might happen. In 2026, more of them are about what election offices, voters and investigators are actually doing with AI.
Less news isn't less risk
Something being in the news less doesn't make it any less worrisome or any less important. Election officials haven't stopped worrying: 63 percent of local election officials in the Brennan Center's 2026 survey said they are concerned about AI making their jobs more difficult or dangerous. Last week, federal officials told election offices to be on guard against AI-powered tricks ahead of the midterms. And OpenAI has said it is readying defenses against cyber and misinformation threats this election season.
If anything, the most important AI stories in elections may now be the ones that never make headlines: a county quietly using an AI tool to draft voter guidance, a chatbot giving a voter the wrong deadline, a fake video that circulates in a group chat and never reaches a reporter.
For election officials, that means planning for the AI risks that won't make the news, not just the ones that did in 2024. You don't have to do it alone:
- Ready for Tuesday offers AI trainings, workshops, tabletop exercises and hands-on help putting AI policies and safeguards in place. We're happy to work with you. Get in touch.
- The AI + Elections Clinic at Arizona State University trains election officials to use AI tools responsibly and to prepare for AI-driven threats.
- The Election Security Exchange is a nonpartisan hub for cybersecurity and physical security resources built for election offices, including guidance on AI.
A tracker you can use, and help build
I've put all 201 stories on a map, by state and year, so you can see where AI has shown up in election news since 2023. I'll keep it updated as new stories come in.
Explore the AI Election News Tracker
Headlines only catch part of the picture, so I want your help. If you know of a U.S. news story about AI touching an election, an election office, or voters that isn't on the tracker, send it in. Local stories are especially welcome, since they're the easiest to miss.