ES
Methodological transparency

How ContrastaLab
works

A claim auditing system that separates facts, opinions and rhetoric sentence by sentence. No ideology. No absolute judgements. Transparent about what it can and cannot verify.

The analysis process

When you enter a text, ContrastaLab processes it in two independent phases: an immediate rhetorical analysis and an optional factual verification available on demand.

1

Segmentation

The text is divided into units of meaning: individual sentences and claims that can be analysed independently.

2

Type classification

Each unit is classified: verifiable fact, opinion, prediction, emotional rhetoric, criminal accusation, insult, speculative comparison or quantitative data.

3

Evidence evaluation

Assessment of whether the claim is supported by verifiable public data, plausible without a concrete source, or cannot be classified as a fact.

4

Rhetorical evaluation

Analysis of language: whether it is precise and honest, uses simplifications or vagueness, or contains fallacies, insults or manipulation techniques.

5

Verification with sources (optional)

For verifiable facts, the user can request a real-time web search. ContrastaLab searches for real sources and updates the evaluation with concrete URLs.

The traffic light system

Each claim receives two independent evaluations: one on the available evidence and one on the rhetorical quality of the language. A sentence can contain correct data and simultaneously use manipulative language — ContrastaLab separates them.

Evidence evaluation

Green — VerifiedSupported by public data, established consensus or an official source found in the search.
Yellow — PartialPlausible or probable, but without a concrete source cited. May be true, but unverified.
Red — Not verifiable or contradictedCannot be classified as a verifiable fact (opinion, accusation or prediction), or evidence contradicts it.

Rhetorical evaluation

Green — Precise languageClear, honest, no apparent manipulation techniques.
Yellow — SimplificationVagueness, moderate hyperbole or mild generalisation that does not constitute a clear fallacy.
Red — Manipulation detectedDirect insult, logical fallacy, unsupported criminal accusation, dehumanisation or conspiratorial language.

Reference sources

For factual verification, ContrastaLab uses real-time web search, prioritising official and institutional sources: World Bank, IMF, ILO, UN Data, OECD, WHO, IEA and national statistical institutes. Media sources are used to locate primary documents, never as proof of truth on their own.

Limits of the analysis

ContrastaLab is a tool to support critical thinking, not a definitive arbiter of truth. It is important to understand what it does not do:

It does not issue legal or judicial judgements about the people mentioned.
It does not determine the author's intention or whether they are consciously lying.
It has no access to classified, private or unpublished information.
It cannot verify very recent facts not yet indexed in public sources.
It does not replace independent journalistic verification or expert human judgement.
The automatic analysis may contain errors — always cross-check.

Ideological neutrality

ContrastaLab takes no political position. It analyses language and evidence using the same criteria regardless of the ideological origin of the text. An insult is an insult whoever says it. A claim without a source is partial wherever it comes from.

Neutrality does not mean relativism. Where established historical or scientific consensus exists — Francoism was a dictatorship, climate change is human-caused according to the IPCC — ContrastaLab applies it as evidence, not as opinion.