This article is Part 2 of a Democracy Works series about the role of AI in the 2026 midterm elections. Today’s post explores why the existing AI landscape erodes trust and poses a risk to voters ahead of this year’s elections. You can catch up on Part 1 here.
American democracy has a trust problem. Recent redistricting efforts and false voting claims have fueled uncertainty about election integrity, undermining a core pillar of democratic stability and degrading the information environment. Robust voter participation relies on trust.
At a time when Americans need trusted election information more than ever, it is critical that AI tools are up to the task. According to a March 2026 survey, roughly half of respondents were very or extremely concerned that AI will discourage voting or spread inaccurate voting information, underscoring Americans’ hesitancy about the trustworthiness of AI.
As Americans continue adopting AI at a rapid pace in the years ahead, voters need trustworthy information delivered through the platforms they actually use. These systems should be designed to be resistant to misuse and capable of delivering accurate, nuanced election information. If AI is to help solve our trust problem rather than contribute to it, we must mitigate ongoing risks and identify solutions that bridge the trust gap.
Understanding AI Risks in 2026
Ahead of the midterms, American voters face several possible challenges related to AI. Manipulated videos, images, and audio are already proliferating ahead of this year’s elections. Experts have previously warned that deepfakes could be used to imitate authoritative sources of information, such as local election officials, potentially disenfranchising eligible voters with disinformation. In this fraught information environment, it is not enough for voters to know whom to trust—they must also know where to find verified information.
Other AI dangers are more technically complex. Researchers at the Center for Democracy & Technology (CDT) have cautioned that AI models may be vulnerable to “algorithmic poisoning" ahead of the midterms. This occurs when malicious election-related content corrupts an AI model’s training data or skews the webpages from which models retrieve information. The risk grows wherever election jurisdictions do not post high-quality, machine-readable voting information online, leaving a “data void” for disinformation to fill. When official sources are absent, AI systems do not go silent—they hallucinate, and voters get authoritative-sounding answers built on whatever the model happened to crawl.
Whether deliberate or not, the shortcomings of AI are already playing out ahead of November. A May 2026 study found that major AI models answered questions related to election procedures or redistricting incorrectly 31% of the time. That is not a failure of any single model—it is what happens when AI systems are not connected to the authoritative election data that already exists.
Bridging the Trust Gap
What can be done to rebuild Americans’ trust in our democracy? The answers are complex, but a healthier democracy starts with an election information infrastructure that is accessible to voters, resilient to manipulation, and grounded in authoritative sources.
Federal lawmakers recently introduced legislation that would prohibit individuals from knowingly sharing false AI-generated content with the intent to prevent others from voting, such as incorrect information about the time and locations of elections. Regardless of political outcomes in Washington, more urgent measures are needed to bolster information integrity ahead of November. No matter how many voters turn to AI for election information directly this year, all Americans stand to benefit from an elections ecosystem that reinforces trust rather than uncertainty.
Democracy Works has spent more than a decade building it—aggregating verified election information directly from the state and local officials who run our elections, and delivering it in the structured, machine-readable formats that AI systems, search engines, and civic tools can use. The most important question for 2026 is not the number of people who will ask a chatbot about where to vote; it is whether the AI ecosystem that voters increasingly depend on—across search results, news summaries, social feeds, and voice assistants—will be built on the authoritative election data Democracy Works already provides, or improvised from whatever happens to be online.
The next post in this series will explore opportunities to enact this vision while evaluating the future of election information infrastructure. What does it take to create an elections ecosystem for the AI age, and how is Democracy Works already building it? Sign up for our newsletter to receive an email with the next post in your inbox.
.png)