Methodology

How Despun collects, analyzes, and presents the news.

1

Source Selection

We aggregate headlines from 48 news sources across a 7-point political spectrum, pulling fresh content from each source via RSS feeds. This ensures broad coverage and multiple perspectives on each story.

LeftCenterRight
Far LeftLeftCenter-LeftCenterCenter-RightRightFar Right
Far Left5 sources
JacobinThe InterceptCommon DreamsDemocracy NowThe Nation
Left8 sources
The GuardianHuffPostVoxMother JonesSlateMSNBCThe Daily BeastSalon
Center-Left9 sources
NPRThe AtlanticCNNNBC NewsCBS NewsABC NewsNew York TimesWashington PostPolitico
Center9 sources
PBS NewsHourAP NewsReutersThe HillUSA TodayBBC NewsAxiosBloombergBusiness Insider
Center-Right3 sources
Wall Street JournalThe EconomistRealClearPolitics
Right9 sources
Fox NewsNY PostWashington ExaminerDaily WireThe FederalistWashington TimesThe BlazeAmerican ConservativeNational Review
Far Right5 sources
BreitbartNewsmaxDaily CallerGateway PunditEpoch Times

How we classify bias: Our classifications are based on AllSides Media Bias Ratings and the Ad Fontes Media Bias Chart. These aren't perfect—individual articles and authors can vary—but they provide a useful baseline for understanding where sources typically fall on the spectrum.

2

Story Clustering

We use AI to identify when multiple sources are covering the same story. This involves:

  • Semantic analysis to match headlines about the same event
  • Filtering out stories only covered by a single source
  • Prioritizing stories with coverage across the political spectrum
3

Analysis Process

For each clustered story, we use Claude (Anthropic's AI) to generate:

Core Facts

A neutral summary of what actually happened, stripped of editorial framing and opinion.

Framing Analysis

How each source is presenting the story—what they emphasize, what language they use, what context they include or omit.

Perspective Synthesis

Why people on different sides of the political spectrum care about this issue and what drives their views.

Debate Classification

Whether the disagreement is about facts (what happened), policy (what to do), or values (what matters).

4

Expert Consensus

When relevant, we identify whether expert consensus exists on the underlying facts. This includes:

  • Scientific consensus (CDC, WHO, peer review)
  • Legal consensus (courts, legal scholars)
  • Economic data (CBO, Federal Reserve)
  • Statistical data (BLS, Census, FBI)
  • Intelligence assessments (FBI, CIA, DNI)
  • Historical record (historians, archives)

Known Limitations

  • !AI can make mistakes. Always verify important claims with primary sources.
  • !Bias classifications are imperfect. Sources can vary in bias by topic or author.
  • !We only analyze headlines and snippets. Full article context may differ.
  • !Not all perspectives are represented. Our source list continues to expand.