How We Score Political Claims
Every truth receipt follows the same rigorous process. Here's exactly how it works.
Step 1: Claim Detection
We identify specific, testable political claims from speeches, press conferences, social media, debates, and interviews. We prioritize claims that are:
- Specific enough to verify (not pure opinion)
- From significant political figures
- Relevant to current public discourse
- Measurable against available data
Step 2: Claim Decomposition
Complex claims are broken into individual, testable sub-claims. For example:
"This action will save American taxpayers over $1.3 trillion"
Decomposes into:
- Does the $1.3T figure exist in official documents? (Yes)
- Does it represent NET savings? (No — it's gross compliance costs only)
- Are benefits included in the calculation? (No)
- What do independent economists say? (Misleading framing)
Step 3: Data Matching
Each sub-claim is checked against primary sources:
- Government databases: BLS, BEA, CBO, EPA, EIA, Census
- Official records: Congressional voting records, bill text, executive orders
- International bodies: IAEA, IEA, WHO, World Bank
- Peer-reviewed research: Academic studies and expert analysis
- Verified reporting: Major news organizations with primary source documentation
We never use social media posts, opinion pieces, or anonymous sources as primary evidence.
Step 4: Scoring
Each sub-claim receives a score. The overall Truth Score is the weighted average.
Score Ranges
| Score | Rating | Meaning |
|---|---|---|
| 🟢 76-100 | TRUE | Supported by available data |
| 🟡 51-75 | MOSTLY TRUE | Substantially correct with minor issues |
| ⚠️ 26-50 | HALF TRUE | Mix of accurate and inaccurate elements |
| ⛔ 11-25 | MOSTLY FALSE | Contains some truth but overall misleading |
| 🔴 0-10 | FALSE | Contradicted by available evidence |
Confidence Intervals
Every score includes a confidence level (0-100%) that reflects:
- Quality and availability of source data
- Whether classified or unavailable information could change the score
- Degree of expert consensus
- Complexity of the claim
Step 5: Publication
The truth receipt is published with:
- The exact claim and who made it
- The overall Truth Score and confidence level
- Breakdown of each sub-claim with ✅ or ❌
- All data sources cited
- A shareable format for social media
Corrections Policy
If you have better data that contradicts our scoring, we want to hear it. Submit evidence and we'll review. If you're right, we update the score publicly and credit you. Truth isn't static — it's updated when better data emerges.
What We Don't Score
- Opinions: "The economy is doing great" (subjective)
- Predictions: "This will create millions of jobs" (unverifiable until it happens)
- Promises: "I will cut taxes" (tracked separately in our Promise Tracker)
We score claims of fact — statements that can be checked against existing data.
Questions about our methodology? Contact us. We're an open book.