What Are Exit Polls and How Should You Read Them During an Election?

Exit polls are surveys of voters, not official vote counts. They can help explain who voted, which issues mattered, and how different groups say they voted, but they cannot by themselves establish who won. To read one responsibly, check who conducted it, which voters it includes, what a percentage’s denominator is, how uncertain the estimate is, and when the data was collected or released.

Illustrative example only: Sam is following a fictional Senate race and sees a graphic saying Candidate A has an edge among younger voters. The candidates, numbers, and race in this example are invented. Sam should treat the graphic as a survey estimate about a particular group, then check its methodology and compare it with official vote returns as those are reported. The same questions help with real election-night coverage.

What does an exit poll measure?

Traditionally, an exit poll is a survey that asks people about their vote and views after they leave a polling place. It is useful because ballots are secret: an official tally reports votes, but it does not show the demographic characteristics or motivations of the people who cast them. Exit-poll questionnaires may ask about a respondent’s choice, views on candidates or issues, and background characteristics. These answers are estimates from a sample of voters, not a count of every ballot.

The design can be broader than interviews outside polling places. Many Americans vote before Election Day or by mail, so current survey projects may combine in-person interviews with phone, text, mail, or online responses. The exact methods depend on the poll. For example, a National Election Pool methodology statement for 2024 described in-person sampling at Election Day and early-voting sites plus a separate multi-mode survey for early and absentee voters. That document describes the 2024 design; it should not be assumed to describe every poll or a later election.

What is different about 2026 election-night polling?

For the 2026 U.S. midterms, The Associated Press describes the AP Voter Poll as a blend of a pre-election survey and traditional exit-poll techniques. Its October 28, 2026 explainer says the project interviews voters by phone and online across the states, with additional in-person interviews at polling places in 15 states on Election Day. The survey is intended to explain voter groups and motivations; AP says its findings do not replace the vote count, and its topline findings on overall candidate support will not be reported. Read the label on the specific graphic rather than assuming every network uses an identical product.

There is a useful 2026-specific detail for comparing coverage: SSRS says The Voter Poll is conducted for several news organizations, including ABC, AP, CBS, CNN, Fox News, MS NOW, and NBC, and that they receive the same data, as described in its November 3, 2026 explanation for NBC News. As a result, the same underlying survey appearing on several networks is not several independent polls agreeing with one another. Newsrooms may analyze or display shared data differently, so check the poll name and provider.

Timing also needs a label. AP’s October 28, 2026 preview scheduled its first Voter Poll findings for 5:00 p.m. Eastern Standard Time (EST, UTC−05:00) on Tuesday, November 3, 2026, equivalent to 22:00 UTC on November 3, with additional findings later that night. This was the schedule for that AP survey, not a nationwide poll-closing time and not a statement that ballots had been counted. The U.S. returned to standard time on November 1, 2026, according to NIST’s 2026 daylight-saving schedule; NIST lists EST as UTC−05:00. For another election or date, check the source’s current schedule and use the time zone and UTC offset for that date.

How should you read an exit-poll graphic?

1. Identify the source and poll. Find the polling organization, sponsor, field dates, and method statement. The World Association for Public Opinion Research guidelines recommend disclosing a poll’s sponsor, researcher, sampling frame, interview locations or modes, sample size, weighting, and precision.
2. Read the denominator. “Candidate A leads among younger voters” is not the same as “Candidate A leads statewide.” Check whether a percentage refers to all respondents, a state electorate, or one subgroup. A subgroup estimate usually rests on fewer interviews than the full sample.
3. Look for uncertainty, not just the point estimate. A percentage is an estimate, not an exact measurement. Sampling error can be larger for smaller subgroups, and nonresponse or other nonsampling issues can add uncertainty. A 2024 exit-poll methods statement explicitly noted that its sampling error varied with sample design and group size, and that nonresponse could increase total error. If two estimates are close, do not treat a small gap as proof of a real difference.
4. Check when the data was collected. Early interviews and later responses may be combined or updated. A graphic released at one time can represent a different mix of data from a later version. Note the timestamp, election, state, and whether the poll includes early or mail voters.
5. Separate explanation from outcome. Exit polls can suggest patterns and reasons people report for their choices. They are not the official canvass. Use the relevant election office’s reported vote totals and the news organization’s stated projection standards to track the result.

How does Sam apply those checks?

In the fictional Senate-race example, Sam sees Candidate A at 54% among voters ages 18–34. Sam first checks whether that figure describes the whole state or only that age group, and how many respondents are in the group. Sam then looks for the survey’s field dates, whether early and mail voters are represented, the source, and any uncertainty measure. The invented 54% is not a statewide result. Even if several channels show the same number, Sam checks whether they are reporting the same survey before counting them as separate evidence.

Sam also distinguishes the poll’s explanation from the race call. A survey may report that a group favored one candidate or prioritized a particular issue; it does not mean every person in that group voted the same way. Nor does a subgroup pattern settle a close contest. Sam waits for official vote totals and clearly attributed projections, which are based on additional evidence and the remaining ballots—not on an exit poll alone.

Exit polls, pre-election polls, and vote counts are not interchangeable

A pre-election poll asks about intentions or preferences before voting is complete, so it can include people who ultimately do not vote. A traditional exit poll interviews voters after they cast ballots, while a modern election-night survey may include multiple methods to reach early and mail voters. A vote count records ballots reported by election officials. Some products use the phrase “exit poll” broadly, so the methodology matters more than the headline.

For the fictional race, Sam should use each source for the question it can answer: surveys for reported voter characteristics and motivations, official returns for counted ballots, and projections only when the publisher explains its basis. Pollsters can estimate patterns from a sample; they cannot turn that sample into an official result.

Quick checklist

• Who conducted and sponsored the poll?
• Which election, location, voters, and dates does it cover?
• Does it include early and mail voters, and how were they reached?
• Is the percentage statewide or for a subgroup, and what is the respondent base?
• What uncertainty or limitations does the method statement disclose?
• Is this a new survey, a later update, or the same data shown by another outlet?
• What do official vote returns show, and has a news organization explained its projection?

An exit poll is most helpful when read as a carefully sampled but imperfect account of voter behavior. In Sam’s hypothetical race, it can add context to the story; only official counting and transparent projection standards can establish how the election result is being determined.