Methodology

Evidence, confidence, limitations — at every step

Our methodology responds to one simple constraint: never claim that a piece of content is "true" or "false". Every step of the process is designed to trace a claim back to its evidence, and to never go beyond what the data actually allows us to conclude.

  1. 01

    Collection

    Data gathering from public, legally accessible sources: websites, blogs, forums, social media via official APIs. No unauthorized collection on restricted-access platforms.

  2. 02

    Detection

    Identification of coordination signals — timing synchronization, amplification networks, content reuse — rather than an evaluation of the content itself.

  3. 03

    Documentation

    Every finding is archived, timestamped and linked to its primary source, with a confidence score and the explicit limitations of the analysis.

  4. 04

    Review

    Internal review before publication: verification of evidence traceability, tone used, and the absence of unsupported conclusions.

  5. 05

    Publication

    Public release as a bulletin, written to be understandable without technical expertise, never naming a culprit beyond what the evidence allows to establish.

Data sourcing

Current phase: simple, legal public sources (websites, blogs, forums, social media via official APIs). Extension to more restricted platforms will only happen through a licensed source or an official API — never through unauthorized access.

Methodological grounding

The term "coordinated inauthentic behavior" has been used since 2017 by major platforms' security teams to describe this type of manipulation. Our approach draws on a framework similar to that of specialized disinformation researchers, such as the Atlantic Council's DFRLab, which distinguishes three mechanisms: structuring inauthentic communities, exploiting platform technical features, and manipulating ranking algorithms.

Editorial safeguards

  • No conclusion goes beyond what the collected evidence allows to establish.
  • Every bulletin explicitly states its methodological limitations before release.
  • Individuals or accounts mentioned are identified based on observable behavior, never an assumption of intent.
  • A systematic internal review precedes every publication.