Evidence-Based Verification

Our Fact-Checking Methodology

We employ a rigorous, multi-stage verification process inspired by leading fact-checking organizations and academic research standards, prioritizing high-quality scientific evidence and source diversity.

Core Principles

Evidence Hierarchy

We follow the established evidence pyramid, prioritizing systematic reviews and meta-analyses over single studies, RCTs over observational studies, and peer-reviewed research over anecdotes.

Source Independence

Industry funding creates conflicts of interest. We prioritize independent research from EPIC-Oxford, EAT-Lancet Commission, PCRM, and peer-reviewed journals without industry ties.

Transparency

Every claim is traceable to primary sources with direct links to research papers, government reports, and authoritative sources. We implement ClaimReview structured data for maximum reach.

Whole Foods Context

We distinguish between whole food studies and isolated nutrient trials. A beta-carotene supplement study doesn't tell us about carrots—context matters in nutrition research.

Evidence Pyramid

Strongest
Systematic Reviews & Meta-Analyses
Pooled analysis of multiple studies (Score: 90-100)
Randomized Controlled Trials (RCTs)
Can establish cause-and-effect (Score: 75-89)
Prospective Cohort Studies
Long-term observational research (Score: 60-74)
Case-Control & Cross-Sectional
Weaker observational studies (Score: 40-59)
Weakest
Animal/Cell Studies & Case Reports
Cannot extrapolate to humans (Score: 0-39)

Our 5-Stage Verification Process

1

Query Analysis & Expansion

We identify topic category, claim type, and controversy level. For nutrition questions involving animal products, we flag for enhanced source diversity. We generate multiple search query variants to ensure comprehensive coverage.

2

Source Discovery & Validation

We search across trusted domains including The Lancet, BMJ, JAMA, Nature, CDC, WHO, Harvard Health, and independent nutrition research institutions. Each URL is validated for accessibility and relevance.

3

Quality Scoring & Prioritization

Each source receives a quality score (0-100) based on study design, journal impact factor, recency, sample size, and citation count. We prioritize high-quality evidence and exclude sources scoring below 40.

Meta-analysis in top journal (2023, n=50,000): Score 100
Large RCT in JAMA (2022, n=5,000): Score 85
EPIC-Oxford cohort study (2020, n=65,000): Score 80
Cross-sectional study (2015, n=500): Score 45
4

Evidence Synthesis

We synthesize findings using AI with explicit instructions to prioritize high-quality evidence, acknowledge uncertainty, distinguish correlation from causation, flag industry funding, and note study limitations.

5

Rating Assignment

We assign one of five ratings following International Fact-Checking Network standards:

TrueFully accurate
Mostly TrueLargely accurate
Partially TrueMixed accuracy
Mostly FalseLargely inaccurate
FalseCompletely wrong

Nutrition-Specific Approach

Nutrition research is particularly susceptible to industry influence. Studies funded by the dairy industry tend to show favorable results for dairy, while meat industry research often downplays health risks. To combat this bias, we:

  • Prioritize EPIC-Oxford study (65,000+ participants including vegetarians/vegans)
  • Include EAT-Lancet Commission reports on sustainable healthy diets
  • Feature Physicians Committee for Responsible Medicine (PCRM) research
  • Reference NutritionFacts.org evidence reviews without industry funding
  • Distinguish whole food studies from isolated nutrient trials
  • Flag conflicts of interest and industry funding

Limitations & Caveats

What We Can Verify

Factual claims with scientific evidence, government data, or documented events—particularly health/nutrition claims, scientific facts, historical events, and policy claims.

What We Cannot Verify

  • ×Future predictions (cannot verify until events unfold)
  • ×Subjective opinions (no objective truth value)
  • ×Moral/ethical claims (involve values, not facts)
  • ×Claims requiring classified information we cannot access

Note: Scientific understanding evolves. A claim rated "Mostly True" today may require revision as new research emerges. Check publication dates and recognize that older verifications may not reflect current knowledge.

Questions or corrections? Use the report function on any fact-check page.