How we research
The goal of this guide is to inform, not to persuade. It covers 806 races and 1,747 candidates on the November 3, 2026 ballot, with 21,409 cited sources — 10,717 of them official government records.
Coverage
New York City, in depth:Governor, Attorney General, Comptroller, every U.S. House district that touches the city, every NYC State Senate and Assembly district, and the city’s ballot proposals. Everywhere else: every U.S. House and U.S. Senate race in the country.
Sources, in order of preference
- Official records — roll-call votes from the House Clerk, the U.S. Senate, the NY Senate and Assembly; bill texts and histories; FEC filings; court and government records. Shown with a green citation.
- The candidate’s own words — campaign sites, press releases, questionnaires, and verbatim interviews or debates.
- Independent news reporting from established outlets.
- Reference works like Ballotpedia — used for basic facts such as ballot status and party lines, not for positions.
We don’t use anonymous claims, opinion columns, or partisan attack material.
Every claim is cited
Each sentence about a candidate carries numbered citations that open the source. If a claim can’t be sourced, it isn’t published. Each race page lists every source it uses and notes what we couldn’t find. Before publishing, a second research pass re-opens every cited source and corrects or removes anything the source doesn’t support.
Issue positions
The guide groups evidence under 15 issue subjects. Candidate statements are shown with citations. Recorded votes are shown as votes on specific bills, with their dates and official sources. We do not treat a vote on a bill as proof of a broad policy belief, or assume a no vote means the opposite of every provision in a bill. This is especially important for legislation that combines many policies. Statements and actions can be read alongside one another without assigning a consistency verdict. Missing evidence means we have not documented it; it does not establish a candidate’s position.
“Compare by issue” filters documented evidence by subject. Candidates appear alphabetically within each race, with the same topics displayed for everyone. Vote-based evidence is labeled separately from other documented positions. We do not weight issues or calculate candidate scores. Missing research is shown as a gap.
“Where I stand” — how the comparison works
On “Compare by issue” you can optionally answer Agree or Disagree to any of the issue statements (for example, “Stronger rent regulation and tenant protections”). Anything you leave as “Not sure / skip” is left out. For each statement you answered, every candidate’s documented position on that same statement is put next to your answer. What a candidatesays and how they voted are compared separately:
- = Same as you — you agree and their documented position supports the statement, or you disagree and it opposes it.
- ≠ Different from you — the documented position points the other way from your answer.
- ~ Mixed record — the documented position is mixed (for example, support for part of a policy). It is not counted as same or different.
- Recorded votes, one at a time — for officeholders, each key floor vote on the topic is listed with its official roll call. For every vote we note which way a Yes points relative to the statement — only when the bill’s central effect squarely matches it (for example, a Yes on the Bipartisan Background Checks Act is in line with “Stricter gun laws”; a Yes on concealed-carry reciprocity is not). Each vote is then marked = In line with your answer or ≠ Not in line with your answer. A No vote is read only as a vote against that measure — never as support for some opposite policy — and we never turn a set of votes into an overall stance. Votes where we haven’t recorded a direction are shown but not compared. The direction notes are in the site’s open-source code, with the provision each one rests on.
- – No documented position — our research found nothing on the topic. That is a gap in our research, not a position.
Each comparison shows the summary, the votes and their citations. Under each candidate we give plain tallies that keep statements and votes apart — for example, “Statements: same as you on 3, different on 1 · Recorded votes: 4 votes in line with your answers, 1 not · No record on 2 of the 7 topics you answered”. There is no percentage, score, meter, weighting, or “best match”. Every answered topic counts the same, candidates stay in alphabetical order, and the comparison is not a recommendation.
Limits.Research coverage varies a lot between candidates — incumbents and well-funded campaigns leave more public record, so a candidate with fewer documented positions will show more “No record” results. Missing evidence is not a position. Statements are not actions, and recorded votes are not broad policy positions. A single statement can’t capture every nuance, and candidates may have changed their positions since the sources were published, so check the dates on the citations. The issue statements are fixed wording and may not match exactly how you would put it.
Privacy.Your answers aren’t stored by us or tied to you. They live in the page’s link (as ?stand=…, so you can bookmark or share it) and, for convenience, in your browser tab until you close it. Like any web address, the link may appear in routine web-server request logs, and anyone you share it with will see your answers and any address it contains.
Congressional votes & money
Key votes are read by script straight from the official House Clerk and U.S. Senate roll-call XML, and matched to members through the public congress-legislators crosswalk. Campaign-finance totals come from the Federal Election Commission’s bulk candidate summary for the 2026 cycle.
State legislative candidate lists
For states other than New York, the candidate lists for state legislative races come from Ballotpedia’s statewide general-election tables (5,382 contests in 45 states, generated 2026-10-19). They are names and parties only, not researched in depth. Ballotpedia notes its lists may be incomplete, write-in candidates are not marked, and seats decided in the primary (Hawaii, Oklahoma) are omitted. We match each contest to your address by district name and to the sitting member by name and district.
Election results
For Congress and statewide offices, a “projected winner” means NBC News has projected that candidate to win, and vote counts are NBC’s. We checked those calls against Ballotpedia and Wikipedia. Ballot measure outcomes and New York legislative results come from other sources, named on each race. Results are unofficial until each state certifies them, and counts can change. We don’t yet have results for state legislative seats outside New York.
Who does the research
Profiles are researched and written by AI research agents working to a written standard (neutral language, primary sources first, verbatim quotes only, never fabricate) and checked by automated validation of every citation. AI makes mistakes. Always follow the citation, and check with your state or county board of elections for the final word on what’s on your ballot.