01 — The Project
ABOUT
An independent research initiative measuring political bias in artificial intelligence.
02 — Mission
Why this matters
Large language models are rapidly becoming the world's default source of information, advice, and even opinion. Billions of conversations happen each day with systems whose political inclinations remain largely unexamined.
GPT at the Polls brings transparency to this critical issue. By presenting AI models with real U.S. congressional votes and comparing their responses to the recorded positions of actual lawmakers, we quantify political alignment in a way that is empirical, reproducible, and open.
Our goal is not to condemn or endorse any model. It is to give the public, researchers, and policymakers the data they need to understand the political tendencies embedded in the AI systems that increasingly shape public discourse.
03 — Process
How it works
Select Real Votes
We curate hundreds of roll-call votes from the U.S. Congress spanning healthcare, defense, immigration, civil rights, the economy, and more.
Query Every Model
Each AI model receives the full bill text and is asked to cast a Yea or Nay vote, along with a justification. Responses are collected via official APIs.
Measure Alignment
We compare each model's vote against those of two reference legislators—one from the left (AOC) and one from the right (Speaker Johnson)—to produce a political alignment score.
04 — Values
Our principles
Transparency
Every prompt, every response, and every scoring criterion is published. Our methodology is fully open so that anyone can replicate our results.
Neutrality
We do not take a political position. The project exists to measure and report, not to advocate. Our reference legislators were chosen for their consistently strong party-line voting records.
Rigor
Data is sourced directly from official congressional records and collected through official AI model APIs. We document limitations honestly and update our analysis as new data becomes available.
Independence
GPT at the Polls is an independent project. We are not affiliated with, funded by, or endorsed by any AI company, political party, or lobbying group.