A push poll is political persuasion dressed up as a survey. The caller claims to be conducting a poll, then asks loaded questions meant to plant a negative idea about a candidate, such as “Would you be less likely to vote for Smith if you knew she was under investigation?” Nobody analyzes the answers. The goal is to change your mind, not measure it.
Push polls aren’t polls at all, so they never produce published results. They’re telemarketing, usually aimed at thousands of voters late in a close race.
How does a push poll work? #
A push-poll call often starts like a real survey: the caller says they’re with a research group asking about an upcoming election. After a question or two, the call turns to “hypothetical” questions built around a damaging claim.
The claim may be false, exaggerated or stripped of context. Framing it as a question lets the caller spread it without stating it outright. Whatever you answer, the call ends quickly, and the operation moves on to the next number. If you’re left thinking “I didn’t know that about Smith,” the call did its job.
Because the aim is reach, push polls go to as many voters as the sponsor can afford, often by automated call. A real poll interviews a few hundred to a couple thousand people selected to represent the electorate.
The most cited example comes from the 2000 South Carolina Republican presidential primary, when voters reported calls asking loaded questions about John McCain’s personal life. Nobody ever released results from those “surveys,” because there weren’t meant to be any.
Push polls vs. legitimate message testing #
Real campaign polls also ask pointed questions. Before spending money on ads, a campaign tests arguments for and against its candidate and opponent. “If you knew Candidate A voted to raise the gas tax, would that make you more or less likely to support him?” can be a legitimate question in a message-testing poll.
The difference is purpose and scale:
| Legitimate poll or message test | Push poll | |
|---|---|---|
| Goal | Measure opinion | Change opinion |
| Number of calls | Hundreds to about 2,000, designed to be representative | Thousands to hundreds of thousands |
| Length | Often 10 to 20 minutes | Often under 2 minutes |
| Demographic questions | Yes, needed for weighting | Few or none |
| Balance | Tests several arguments, often on both sides | One-sided attacks on one candidate |
| Who’s calling | Research firm named on request | Vague or evasive about sponsor |
| Results | Analyzed; sometimes published | Nothing to analyze |
Message testing and mid-campaign brushfire polls are part of normal campaigning. A push poll borrows their format to deliver an attack.
How can you spot a push poll during the call? #
Look for more than one of these:
- It’s very short. A real survey asks about your age, education and other basics so the results can be weighted. A call that skips all of that and heads straight to damaging “questions” isn’t collecting data.
- The questions are one-sided. Repeated “If you knew…” or “Would it change your mind if…” questions about only one candidate are a red flag.
- The caller won’t say who’s paying. Legitimate firms identify themselves and, if asked, will usually tell you who they are or give contact details. Push-poll operators dodge.
- There’s no room for “not sure.” Real surveys let you say you don’t know. Push polls keep you boxed into their framing.
- Everyone you know got the same call. Scientific polls reach a small random slice of voters. If your neighbors and relatives in the same district all got an identical call the same week, it’s a mass contact program.
Are push polls legal? #
In the U.S., push polls are generally protected as political speech under the First Amendment, so there’s no federal ban on the practice itself. A few states, New Hampshire among them, have laws requiring push-polling calls to disclose whom they’re made on behalf of. Robocall rules also apply: the Federal Communications Commission ruled in 2024 that robocalls using AI-generated voices fall under federal restrictions on artificial voices, after a fake robocall imitating President Biden’s voice targeted New Hampshire primary voters.
The polling profession treats push polling as an ethical violation. The American Association for Public Opinion Research condemns it and has criticized organizations that use it.
What should you do if you get a push poll? #
- Hang up. You don’t owe the caller anything, and staying on the line only gives them more time.
- Write down what you can: the number, the date and time, the candidate named, the claims made and any organization name given.
- Report it. You can send details to your state attorney general or election office, to AAPOR, and to the campaign that was attacked. For robocalls, you can also file a complaint with the FCC.
How do you find real polling instead? #
Real polls publish who conducted them, when, how many people they reached and how. Election Tracker lists published polls by race, each with pollster, sample size, population and field dates, next to a 30-day average. A push poll never appears there, because it never produces results to publish. Our guide to telling if a political poll is reliable covers what to check in a poll you do see, and who pays for political polls explains sponsorship.
Frequently asked questions #
What is an example of a push poll question? #
“Would you be more or less likely to vote for Senator Jones if you knew he voted to cut police funding by 20%?” asked of thousands of voters, when the vote in question was actually a broad budget bill. The question plants a claim; the answer is irrelevant.
Is a push poll the same as a negative poll question? #
No. Legitimate polls sometimes ask about negative information to test how voters react. What makes it a push poll is the intent to persuade at scale, the lack of real data collection and the concealment of who’s behind it.
How is a push poll different from a snap poll? #
A snap poll is a real survey fielded quickly after an event like a debate to measure reactions. A push poll isn’t a survey. Our guide to snap polls explains how rapid-response polling works.
Can I tell a push poll from a text message? #
The same signs apply to texts that link to “surveys”: one-sided questions, no clear sponsor and no demographic questions. Be careful with links in unsolicited political texts, and don’t enter personal information you wouldn’t give a stranger.