Data for Good Bulgaria and Pro Veritas (За истината) are starting MAP-Desinfo, a project focused on monitoring and analysing harmful narratives in Bulgaria.

The project’s duration is from May 2026 to July 2027. It is funded by the Partnership Fund (PACT), administered by the Open Society Institute – Sofia.

The aim is practical: to build a workflow that helps researchers and journalists track how harmful narratives appear, spread and change across different media environments. This includes traditional media, online news sites, YouTube, TikTok, Facebook and Instagram.

Why We Are Doing This

Public debate in Bulgaria is affected by recurring narratives about the EU, NATO, Russia, migration, gender, elections, the judicial system and other sensitive topics. What matters is not only whether a single claim is true or false, but what kind of story it builds: who is presented as a threat, who is blamed, who is cast as a victim, and what reaction is encouraged. This framing can be harmful, especially when it is repeated across social media, online media, public figures and local contexts.

At the moment, much of this work is still done manually. Journalists, fact-checkers and researchers often track cases one by one. That remains necessary but it is hard to see broader patterns this way.

MAP-Desinfo is meant to support that work with a more systematic process. The project will combine media monitoring, machine-assisted classification and journalistic analysis. The result should be a clearer picture of which narratives are active, where they circulate, and how they are framed.

What the Project Will Build

The project is organised around a simple pipeline.

First, we will collect relevant media content from monitoring systems (such as Sensika) and public online sources. This includes text, images and even transcripts from audio and video content.

Second, we will use cost-efficient language models for the first pass through the collected material. They will help separate relevant from irrelevant content, so that manual review can focus on the cases most likely to show recurring patterns.

Third, we will analyse the filtered material in more depth using more capable frontier models. These models will map shortlisted articles to pre-identified patterns of manipulative content and produce detailed reports on how harmful narratives are built and how they spread.

The project will also develop a dashboard for exploring the collected and analysed material, helping journalists track what is circulating in their information space.

The Role of Journalists

Pro Veritas will coordinate a group of regional journalists who will document concrete cases from their areas of work. Their role is essential because the regional information environment is often hard to read and misunderstood by national media.

National media monitoring can show what is visible at the centre, but it often misses how narratives are adapted to a specific town, community, minority group or local conflict. Regional journalists are better placed to notice these local versions and explain their context.

Their work will also help improve the methodology. The project will not rely only on predefined categories. As journalists encounter new cases, the narrative list and classification process can be refined.

Methodology First

Before the monitoring system can be useful, the project needs a shared methodology. We need to define what is defined as a harmful narrative in this project, how cases are documented, and how automated classification should be checked.

The early phase will focus on this groundwork: identifying initial narrative categories, designing the journalist reporting format, and testing how well automated tools can support the analysis.

What Comes Next

The next step is to build the methodology, the technical system and the workflow with journalists. After that, the project will move into regular monitoring and analysis.

We will publish updates as the work develops, including information about the methodology, the dashboard and the analytical outputs.

MAP-Desinfo is implemented by Data for Good in partnership with Pro Veritas as part of the “PACT” project and is funded by the European Union and the Open Society Institute – Sofia (OSIS).🇪🇺