Methodology

How does Vaperina Fact work?

Vaperina Fact is a science-based fact-checking platform specialized in vaping. Our goal is to examine the accuracy, context, and strength of scientific evidence behind claims.

1. Analysis process

Submitted articles go through a two-step process before appearing on the site:

AI analysis

Anthropic's Claude model identifies the article's key claims, confronts them with scientific literature using scientific databases and DOI-referenced studies, and returns the result in structured JSON format.

Human editorial review

Every AI-generated analysis is reviewed by a human editor before publication. If the analysis is inaccurate, contains incorrect sources or reaches misleading conclusions, the editor corrects or rejects it.

2. Verdict system

Every claim and article is assigned one of three categories. Important: these are scientific assessments, not legal findings.

Misleading

The claim is not sufficiently supported by science, omits important context, or suggests an incorrect conclusion.

Mixed

Some claims are scientifically sound, others are misleading or lack context. The article as a whole gives a mixed picture.

Accurate

The claims are consistent with the best available scientific evidence and discuss the topic in appropriate context.

3. Source reliability score

We calculate a reliability score for each media outlet based on analysed articles. The score ranges from 0 to 100.

Formula
score = ((accurate × 2 + mixed × 1) / (total × 2)) × 100

60–100

Reliable

The majority of analysed articles are scientifically accurate.

35–59

Mixed

Mixed reliability – contains both accurate and misleading articles.

0–34

Low reliability

The majority of analysed articles are misleading or lack context.

The score is only shown when at least 3 articles have been analysed. With fewer articles, "Limited sample" is displayed – this reflects methodological caution, not a judgement.

The score is not a final assessment of the outlet’s entire editorial work and is based solely on articles analysed by the platform.

4. Scientific principles applied

Relative harm: Where relevant, we primarily compare e-cigarette claims to conventional cigarettes, taking risk reduction into account.

Cell culture ≠ human evidence: We treat in vitro and animal study results with caution and do not consider them direct human evidence.

Regulated vs. unregulated products: We distinguish between EU TPD-compliant regulated products and illegal/black market products.

Dual use ≠ exclusive use: The risk of combined smoking and vaping is not the same as vaping alone.

Priority of scientific consensus: Systematic reviews, meta-analyses and large cohort studies take precedence over individual studies.

Multi-sided approach: We consider both harm reduction perspectives (e.g. UKHSA, Cochrane) and precautionary principle perspectives (e.g. WHO).

5. Corrections and updates

If an analysis contains an incorrect finding, please let us know. We review all correction requests within 15 working days. Where justified, we update the analysis.

[email protected]