This post is the final installment of Democracy Works Insights, our series tracking voter engagement through the 2026 midterms. Drawing on real-time data from our voter tool TurboVote and our Elections API, we shared insights election by election and state by state, as part of our mission to help every voter make a plan to vote and vote with confidence.
Between September 30 and November 2, Democracy Works sent 13.5 million TurboVote election reminders to voters in all 50 states and Washington, D.C. Each reminder linked to a personalized Election Guide with a voter's sample ballot, their voting location, and the key dates they needed to make a plan to vote. We also helped millions of voters access reliable election information through our data partnerships with AI platforms including ChatGPT, Claude, and Gemini.
Over five weeks of Insights, we watched voters move through making their plans to vote in real time. Here's what the full picture shows.
Voters build their plan in steps
When our first reminders went out on September 30, 14.5% of engaged TurboVote subscribers looked up where to vote. By our final reminder on November 2, the night before Election Day, that number was 58%. Voters were four times more likely to look up where to vote than they were five weeks earlier. On average, 44% of engaged TurboVote subscribers researched where to vote in TurboVote.
Interest in researching candidates and measures on the ballot was not so variable. About a third of engaged subscribers researched what's on their ballot throughout the cycle: 34% on September 30 and 32% on November 2.
The pattern is consistent: voters first learn what they're voting on, then figure out where and how, then act. Accurate election information has to meet them at each step.
Young voters made the biggest leap
On September 30, just 9% of engaged TurboVote subscribers ages 18–24 looked up where to vote. On November 2, 66% did, about seven times the rate. That was the highest of any age group in our final reminder. Across the full cycle, young voters were 18% more likely than subscribers overall to look up where to vote (52% to 44%).
Many young voters are casting a ballot for the first time, or voting from a new address after a move or the start of a school year. For them, finding a polling place can be one of the biggest unknowns. TurboVote partners with over 90 colleges and universities to ensure that student voters can have access to the tools they need to get registered, check their ballots, and look up their voting locations all in one place.
Older voters were prepared from the start
Subscribers 65 and older were the most engaged age group researching election information all cycle long. Across all of our reminders, 42% of engaged subscribers 65+ researched what's on their ballot, compared with 32% of subscribers overall. They also shared their Election Guide at nearly twice the average rate. We first noticed this trend in New Hampshire: digital tools are an effective way to reach and support older voters.
Trusted messengers made a difference throughout the cycle
Throughout the midterm cycle, one pattern kept appearing: voters engage more deeply when election information comes from an organization they already know. In October, we saw it with nonprofit partners in Texas and employers in Kansas. The final data confirms it.
- Nonprofits: Engaged subscribers who signed up through a nonprofit were 32% more likely to have voted than subscribers overall.
- Labor unions: Subscribers who signed up through a labor union shared their Election Guide at twice the average rate.
- Government partners: Subscribers who signed up through a government partner were 41% more likely to look up where to vote than subscribers overall.
We can't say the partner connection alone caused these differences. Voters connected to these organizations may already be more engaged. But the pattern is clear: TurboVote partners, from nonprofits to unions to election offices, help turn intention into action.
Hispanic/Latino subscribers reported voting at the highest rate
In our final reminder, engaged Hispanic/Latino subscribers were 27% more likely than subscribers overall to self-report that they had already voted.
Asian and Pacific Islander subscribers were the most likely to look up where to vote, at 62% on November 2. That echoes what we saw in Texas in our very first Insights post.
The data doesn't tell us why, and differences in how and when each state votes may play a role. But it underscores why polling place information needs to be accurate, easy to find, and available in the languages voters speak.
Voters chose the option that fit their state
Among subscribers who told us how they voted, about half voted early in person. The voting options available in a state shaped the information voters looked for in TurboVote. In Colorado, California and Oregon, where every registered voter receives a ballot in the mail, engaged subscribers were among the most likely to report they'd already voted by November 2. They were also among the least likely to look up a polling place: 29% in Oregon and 43% in Colorado, compared with 58% overall.
Toss-up races weren't the whole story
High-profile races draw headlines, but they didn't predict how engaged voters would be. In New Hampshire, 49% of engaged subscribers researched what's on their ballot. West Virginia (48%) and Wyoming (46%) were close behind but had no statewide race rated competitive. In our final reminder, Indiana led the country in looking up where to vote, at 70%. And in Louisiana, 8.9% of engaged subscribers shared their Election Guide in our final reminder, nearly twice the national rate.
Every voter deserves clear, accurate information, not just those in states that make the news.
Voters turned to AI for election information
TurboVote is only one place voters look. Through our Elections API, Democracy Works also delivers accurate, nonpartisan election data to integrated AI platforms. From October 1 through Election Day, those platforms generated an estimated 25 million requests for voting information covering all 50 states and D.C.
More than half of those requests came in the final week. On Election Day alone, we fielded more than 3 million requests, peaking at 30 requests per second at 2 p.m. ET. Significant activity came from states with closely watched races, including Alaska, Michigan, Texas, Kansas, Iowa, Maine, and South Carolina.
Wherever voters look for answers, those answers should come from accurate, verified sources. That's what the Elections API is built to provide.
What we learned
Five weeks and 13.5 million reminders point to three lessons for anyone helping voters get ready:
- Meet voters at every step. Ballot information matters throughout the cycle; voting locations matter more every day as Election Day nears.
- Trusted messengers deepen engagement. Nonprofits, unions, employers, and election offices help voters act on the information they receive.
- Every state matters. Engagement was strong in states with and without headline races, and every voter deserves accurate information about what's on their ballot and where to vote.
We'll carry these lessons into how we build TurboVote and the Elections API for the elections ahead. Thank you for following along.
Support this work
Our work to build voter confidence is only possible with support from people who believe in the importance of accurate election information, everywhere voters go looking for it. Reaching voters with the right information at the right moment takes resources. If this series has been valuable to you, consider making a gift to Democracy Works today to help us keep reaching voters through every election ahead.
About the data
All TurboVote figures reflect engaged subscribers (those who opened their personalized Election Guide) who received election notifications between September 30 and November 2, 2026. "More likely" figures compare rates between groups (for example, 13.7% is 32% higher than 10.4%). "Already voted" and voting-method figures are self-reported by subscribers and are not verified turnout data. Age and race figures are national and reflect subscribers matched to demographic data. Elections API figures reflect data requests from integrated AI platforms and do not correspond to individual user conversations. Analysis shows differences in behavior, not cause and effect.
.png)
.png)
.png)