Methodology

Individual articles and episodes are rated by our team of over 40 trained analysts. Analysts work in three-person politically balanced pods — one leaning left, one center, and one leaning right — and conduct live panel discussions via Zoom. Each analyst rates independently, and scores are averaged to produce the final overall rating for a source.

We rate all article types — news, analysis, and opinion — using an independent classification methodology, regardless of how the source itself labels the content. This allows us to capture the full range of a source's output rather than relying on self-reported categories.

We acknowledge that everyone and everything carries some bias — including us. That's why we've built explicit bias-mitigation processes into our methodology: balanced analyst pods, rule-based scoring criteria, ongoing training, and regular calibration sessions. Our detailed white paper explains each mitigation technique in full.

Analysts sample from prominently featured articles across multiple news cycles, which allows us to evaluate bias-by-omission alongside content quality. Longer-running sources accumulate larger sample sizes over time. Same-day pulls are used whenever possible to capture real-time editorial decisions.

Human analysts create and maintain the primary Media Bias Chart®. AI is used to augment human ratings for business customers, enabling real-time content analysis at scale. AI-assisted ratings are always derived from and validated against our human-generated baseline.

We update sources daily across the entire database. The most widely-read and widely-watched sources are updated most frequently. We continuously archive rated content and add new sources as they become prominent. This rolling update process keeps ratings current and representative.

Analysts work live shifts inside our proprietary CART (Content Analysis Rating System) platform. Each piece of content is scored across six sub-factors:

Each sub-factor has precise definitions and criteria. Analysts require a minimum of 20 hours of pre-rating training, including live practice sessions with calibrated examples.

Founder Vanessa Otero created the original methodology in 2016 as a solo project. It evolved steadily through input from academics, journalists, and domain experts. The first multi-analyst project was completed in June 2019, and analyst teams expanded significantly through 2020. Today we employ 40+ paid analysts who rate content daily.