Since 1848, The Associated Press has played a unique – and essential – role in U.S. elections, counting the vote and declaring winners in races across the country. Ahead of the 2026 midterms, David Scott, vice president of elections, explains how and why AP counts the vote, why results may fluctuate on election night and the extensive steps AP takes to ensure accuracy.
AP has counted the vote and called elections in the U.S. since 1848. Why does AP play this role in the first place? Why is this the job of a news organization?
For more than any other reason, AP counts the vote because it’s news. There’s no bigger story on election night than who Americans have picked to lead. At the dawn of the cooperative, a desire to break that news led AP to count the vote for the first time.
We are able to break that news, to tell that story, because of the way the founders of this nation directed that elections for federal office be conducted. They kept their instructions in the Constitution brief and to the point. For Congress, for example: “The Times, Places and Manner of holding Elections for Senators and Representatives, shall be prescribed in each State by the Legislature thereof…”
The choice to leave the mechanics of Election Day to the states means every state does it differently. Perhaps the only thing every state has in common is that no state runs elections at the state level. Elections in America, even those that take place in November midterms for national office, are a collection of local elections – administered by tens of thousands of local election officials. To be the first to find out who has won the most votes, you need the results from every county, or parish in Louisiana, or town in New England.
In this year’s midterms, that means we’re getting votes from more than 5,000 “reporting units.”
That’s a big number, but that’s what we do and what we have done for 178 years. AP counts the vote. Not officially, of course. The actual work of tabulating results ballot by ballot is the job of election officials. But once polls close, to find out who Americans have picked to lead, you’ve got to do the work of collecting those results town by town, county by county, and parish by parish. You’ve got to count the vote.
What, if anything, is different about this year’s midterms compared with 2024 or previous elections?
To start, we’re really excited to again be providing election results to all of the major U.S. television networks – ABC, CBS, CNN, NBC and MSNOW rejoining Fox as AP Elections customers. Given that voters continue to be confronted with misinformation, we think it’s essential that news organizations have one source for the data they need to serve voters with accurate and trusted results.
This will also be the first major general election in which we’ll be using The Voter Poll. That’s our new survey of the American electorate, which is a combination of traditional exit polling and the methodology that AP developed and used with AP VoteCast. Rather than running the survey ourselves, we’re working with the U.S. television networks to conduct that research.
Another big change is in the nature of our customers. The world’s leading technology and artificial intelligence companies are among our biggest customers, right there alongside the networks, major international broadcasters and local television and newspaper ownership groups.
Those tech firms have an audience, not unlike the one we have as a global news organization and our media customers have as publishers and broadcasters. They see the value of the election story and want to be a part of it, and we’ve worked hard to make sure they get from AP what they need to tell it.
It’s all part of what the AP Elections team calls the “elections ecosystem” – the idea that no matter where you look for news on election night (and in the days after), those telling the story or serving up the data are depending on AP for the most basic question: who has got the most votes.
Something we all see on election night is fluctuating real-time results as the votes come in. Why might a candidate’s vote total change – or even decrease – as results are reported on election night? And why might a large number of votes appear to be added to the results at once?
Most ballots cast in American elections are still marked-up pieces of paper. There will be well over 100 million of them in this year’s general election. Voters are human – and so are the people who collect and count their ballots. They make mistakes on occasion. Those mistakes are almost always caught within a few minutes. Sometimes within a few seconds.
Not every change in the vote count is the result of a mistake, of course. Election officials often count and report results in batches, which means thousands of votes can be added to the totals all at once. That’s a normal part of the process.
I understand why people might worry about this. I’m a journalist and for close to 30 years have lived by AP standards, which put accuracy in reporting the news at the top of our list of values.
But I’ve also been doing this work in elections for more than a decade. What I have witnessed – year after year – are election officials, and our own AP Elections team, taking on what is honestly an impossible task and – year after year – succeeding at delivering the impossible.
What does AP do when it sees a vote total or other election data that appears unusual or inconsistent? How does AP distinguish normal changes in the results from something that warrants further scrutiny?
To start, we’ve got an entire team on election night focused on what we call “quality control.” A simple and common term for what is an enormous effort. There are a few things I’d highlight.
First, just like any other AP story, we always want to have more than one source. We want to see that results reported by local election officials to AP match those they’ve posted to their website or placed into their electronic feed.
Second, we look closely at results before they’re published. For example, when a county starts to post results online, we don’t immediately put them into our vote count. We take a few extra minutes to verify that those results look OK: Are we sure they’re actual votes and not test data? Is our system ingesting them properly, so that the votes for a candidate on the website are landing in our database correctly as being for that candidate?
Are there more votes in this update than we expected, and if so, should we slow down and check to see if they or our expectations might be incorrect?
We talk a lot about the accuracy of our race calls. But that rate of accuracy (99.9% so far this year) depends on a vote count that’s just as precise.
Anything else that you think is important to voters to understand how election results change or fluctuate on election night?
I would ask people to be patient. I know that’s so hard to do on election night, when we’re all watching TV or looking at our phone, refreshing apnews.com constantly because we want to know who has won.
But just think about the scale of what is happening after polls close! In just a few hours, election workers will open and tabulate more than 100 million ballots across thousands of races for even more candidates.
In a few hours, we’ll know what direction voters have set forth for the country. If it’s a decisive win, we might know the outcome of most races on election night. If it’s a close contest, it might take a few days. And that’s OK!
It takes time to get things right. The best things always do. And it’s time well spent when it comes to something as important as making sure the voice of every voter is heard via their ballot.
That’s really what the numbers in the vote count are: each number is the voice of an American saying this is what I believe, this is who I want to represent me, this is who I have picked to lead.
Related articles
