Every campaign email CampaignRadar receives is analyzed within seconds of arrival using a hybrid scoring system — combining rule-based detection with AI analysis to produce a single letter grade and detailed breakdown.
We believe voters and journalists deserve to know not just what candidates are saying, but how they say it — and whether the tactics they use to communicate respect or manipulate their audience.
This page explains exactly how we arrive at every score. Nothing is a black box.
CampaignRadar subscribes to campaign email lists across every congressional district using a purpose-built subscription bot. Each candidate is assigned a unique inbound email address — for example, cr-000004@inbound.campaignradar.io — so every email can be attributed to exactly the right campaign.
When a campaign sends an email, it arrives at our inbound mail server within seconds. The email is parsed, stored, and queued for scoring automatically. No human reads the email before it is scored — the process is fully automated from receipt to grade.
We only subscribe to publicly available campaign email lists. We do not intercept or access any private communications. Every campaign we track made a conscious decision to send mass emails, and CampaignRadar receives them as any member of the public would.
Each email is evaluated across four dimensions, weighted to reflect what matters most for voter-facing communication: whether the email informs, whether it is honest about who sent it, and whether it respects the reader.
Campaigns face real, legally mandated filing deadlines with the FEC — quarterly reports, pre-primary disclosures, and year-end filings. An email sent two days before a genuine FEC deadline with urgent fundraising language is meaningfully different from the same email sent mid-month with no deadline in sight.
CampaignRadar cross-references the date each email was received against a database of real FEC filing deadlines and end-of-quarter dates. Urgency language is penalized on a sliding scale based on how close a legitimate deadline actually is.
Rule-based scoring is fast and consistent, but it can miss nuance. A campaign email that avoids every keyword on our manipulation list could still be subtly misleading. That is why every email also receives an AI analysis pass using Claude, Anthropic's AI model.
The AI evaluates each email with full context — understanding tone, intent, and purpose in ways that keyword detection cannot. It also generates the plain-language summary that appears on every candidate profile: a one-sentence explanation of what the campaign is actually asking for and why.
A candidate's overall grade is the weighted average of their individual email scores. New emails update the grade immediately. A candidate who starts with poor communication habits can improve — and one who starts strong can see it fall.
CampaignRadar grades how candidates communicate — not what they believe. We do not score, evaluate, or express any opinion on a candidate's policy positions, political ideology, party affiliation, or fitness for office.
A progressive Democrat and a conservative Republican can both earn an A. A candidate we personally agree with can earn an F. The scoring system is blind to politics — it only evaluates communication quality.
We also do not score based on how often candidates send emails, how much they raise, or whether their campaigns are successful. A candidate who sends one excellent email scores better than one who sends twenty manipulative ones.
If you believe a score is in error, contact us. We review disputes and will correct genuine errors in our data.
Browse candidate profiles across every congressional district — grades, email histories, and plain-language summaries of every message we've received.