How we research

1. News data. News counts come from GDELT: English-language articles from US sources that match our Democratic House query. That isn't all US media. Matches are keyword searches, not human-reviewed stories. Topics overlap, so their percentages can't be added, and each percentage uses its own denominator. A day counts only once its full snapshot is complete; missing or failed days show as unavailable, never as zero. Dates are when GDELT found an article, not verified publication dates. Differences between coverage and candidate positions don't show bias, accuracy or motive. Candidate figures count people with reviewed evidence on a topic. A mention isn't always a specific proposal.
2. The dial. The face on the dial is a caricature; Bill Maher has not endorsed this site. CALM and PANIC are satirical labels, not measured public sentiment.
3. DSA Scare-O-Meter count. Counts Democrats on the November 3 ballot whose membership in, or endorsement by, a group the meter tracks is documented in an official group, House or campaign source. Each group is counted and listed separately; a candidate named by more than one group counts once in the total. Only national endorsements count; chapter endorsements are not included. Endorsement lists are published in full, but membership is counted only where a source names it: we have not yet searched every candidate for membership, so the total is a minimum: research does not yet cover every candidate. Current affiliations aren't projected back in time.
4. Timing. Figures as of Oct 15, 2026, 4:25 p.m. UTC. Method and query version: atlas-gdelt-1.0.0.