Guiding Principles

A candidate's position on Choose Between comes from what that candidate has said or done in public, never from what we think they believe.

Every position is dated and linked to its source, and each one carries a marker. "Own words" means the candidate said it in the first person, in a debate, an interview, a speech or a statement on their own campaign site. "Cited" means we're pointing at an act when their words on that question aren't on the record: a vote, a bill they sponsored or signed, an official statement or a show of hands at a debate. Own words come first; a cited act is used when it's the clearest evidence there is. Third-person descriptions of a candidate don't qualify, and neither do scorecards or ratings from outside groups. The sources behind every position in your matchup are listed in your Compatibility Report, with the marker, the outlet, the date and the link.

We use software, including AI tools, to help find and sort through sources, and a member of our team reviews each position before it goes live. Where a summary on the site was written with AI, it's labeled, and the sources sit underneath it so you can check our work.

A matchup usually features two candidates. Other candidates on the ballot are listed with their party, but their positions aren't compared. Which races and candidates we feature is our editorial call, and leaving someone out says nothing about them.

We don't run one national quiz and apply it everywhere. The issues that separate two candidates in Ohio are rarely the issues that separate two candidates in Arizona, so each race gets questions written for it.

A question makes a quiz only when we can place each featured candidate on it from the record. If we can find one candidate's position but not the other's, the question doesn't run. The questions reach across the policy areas that decide most races: the economy, healthcare, safety, the environment, how government works and education. The mix inside those areas follows what the record in that race supports. Most are questions where the candidates disagree, because those tell you the most; a few are places where they agree.

Every question has to be answered. There's no skipping, so every result in a race rests on the same full set of questions, and nobody's result is tilted by what they chose to leave out. A few questions near the start ask about you rather than about an issue, such as your ZIP code, which tells us your state. They play a small part in your result and are kept apart from the policy areas.

The questions, their order and their topics can change as the record changes.

Your match with one candidate never depends on your match with another. When you finish, we score every candidate in the matchup against your answers separately.

On each question, a candidate earns credit when your answer lines up with their recorded position. Questions don't all carry the same weight: one that truly divides the candidates in that race counts for more than one where they mostly agree, and the weight shifts a little with how much that issue tends to matter in your state. Your match with a candidate is the credit they earned out of the credit available, shown as a percentage.

Here's the arithmetic on a made-up race between Alvarez and Brooks, with a quiz of three questions and a reader who answers Yes, No, Yes.

The first question carries the most weight because it's the one that pulls these two candidates furthest apart. Both earn credit on the third because both agree with the reader. Real quizzes have more questions and finer weights; the arithmetic is the same.

Your report shows more than the percentage. It breaks the match down by policy area so you can see where you and a candidate line up and where you part ways, and you can move between every candidate in the matchup.

We never ask your name, and we can't connect your answers to you.

Your answers are tied to a random code stored in a cookie in your browser. Clear the cookie and they're gone from your browser; we can't bring them back, and we can't find them for you either, because nothing links them to a person. Combined with many other people's, your answers become anonymous totals: how a race's matches are splitting, which way that split is moving, how people in a state lean on an issue. Those totals feed the figures on the site and the data products we offer to media organizations and campaigns. They're never small enough to point at anyone, and no client ever sees an individual's answers.

We don't sell personal information and we don't share it for targeted advertising. Sharing a result is your choice, and anyone you share it with can pass it on. The Privacy Policy is the full account; if it and this document ever differ, the policy wins.

Choose Between is free, and it's independent.

Nobody involved in running the site works for, is paid by or is affiliated with any candidate, campaign, party, committee or interest group. Three things pay for the site, and none of them can touch the questions, the positions, how results are calculated, the figures we publish or how we handle corrections.

The first is paid political advertising. Those ads are labeled "Paid political advertisement" and kept apart from candidate positions and from your results. Every buyer of a paid position in a race is offered the same terms, and we review each ad before it runs. The second is sponsored content that isn't political, labeled as sponsored. The third is the data products described above, sold under terms that forbid any attempt to identify anyone. An ad on the site isn't an endorsement from us, and it doesn't mean the advertiser endorses us.

Candidates change their minds, sources move, and new statements land after we publish.

We work to keep positions accurate and current, but we can't promise every one is complete or up to date at every moment, so check the linked source before you lean on it.

If you spot a problem, use "Request a correction" on any candidate profile and tell us the race, the candidate, the position you think is wrong and a public source that supports the change. Readers, candidates and campaigns are all welcome; campaigns should write from an official address. Once a correction checks out, we update the page and usually add a dated note saying what changed. If a candidate disagrees with a position that's accurately sourced, we may add a dated note from them beside it rather than remove it. We don't take down an accurate, sourced position because someone asks, though a candidate can raise it again with a new source while the race is live.