The Bulgarian parliamentary elections have been a long-distance runner’s game since 2021 with the latest checkpoint being on October 27th, as over a quarter of the population went to the ballots for the seventh consecutive time in the last three years.

As the crosshairs of the new wave of Russian foreign information doctrine have become an odd comfort zone, the media and communication spheres have been utilised in sowing discourse and apathy which ultimately wanes popular participation in the democratic process.

To gain ongoing insights into the information flow, we have been collaborating with media and communication experts, disinformation and misinformation analysts and decision-makers to provide them with an actionable 360-degree perspective.

These tools are combined in our annotation and exploration platform—Identrics’ Annex—which we utilised ourselves to put together this sitrep.

What we analysed for this report

Data

We are looking at

  • online news media and news agencies,
  • blogs, forums and message boards,
  • government, regulatory and organisational sources,
  • social communication networks, as usual origin points for coordinated propaganda campaigns,
  • as well as traditional media in newspapers, television and radio.

Mechanics

We are using

  • proprietary language models to analyse narrative spread across datasets,
  • natural language processing to understand trends,
  • information extraction to pinpoint relevant entities,
  • and engagement measurement to understand the narrative impact,
  • as well as proprietary models for synthetic propaganda,
  • and hate-speech detection.

Scope

We read this article, which is informed by this poll. We used this list of parties and official party leaders to design the focus points of this sitrep.

The 180 degrees of media intelligence:

Context and media sentiment before the elections

To analyse the narrative in detail, we captured both the general sentiment and the historical background leading up to the elections.

We used traditional media intelligence to tell us the “who,” by measuring entity mentions in information mediums and gauging overall sentiment. This revealed the complex power struggles and alliances that shaped the election discourse:

Mentions of official party leaders across the infosphere.
  • The conversation leading up to the elections was dominated by the discord of Delyan Peevski’s Movement for Rights and Freedoms—New Beginning (MRF-New Beginning) coalition and Ahmed Dogan’s Alliance for Rights and Freedoms (ARF), along with their gravitation around Boyko Borissov, the leader of GERB.
  • The dramatic split within the Movement for Rights and Freedoms (MRF) reshaped the coalition dynamics. The consequences became apparent in the results, which led the reformist coalition We Continue the Change—Democratic Bulgaria (WCC-DB) and the pro-Moscow party, Vuzrazhdane, to second and third places in the elections.
  • With trailing results for the Bulgarian Socialist Party—United Left (BSP – United Left) coalition, Slavi Trifonov’s populist party, There Is Such People (ITN); little over the 4% parliamentary barrier for the nationalist-populist MECH and a hairline below it for the far-right Velichie, which managed to go over back in June of this year.
The general narrative sentiment progression across the infosphere from Monday, October 21st, until its negative peak on Sunday, October 27th, on the day of the elections.

The influence of fast media in the elections

While traditional outlets played a significant role in official campaigning, fringe voices gained traction through fast media like Telegram, Reddit, and TikTok. The far-right “Velichie” party was particularly effective in leveraging these platforms, with over 20% of leader mentions originating from social networks—more than double the average. Even influencer-driven campaigns were gaining traction this time.

While this fast media venue of communication did not translate to electoral success as seen in the last elections—due to larger political shifts—it signifies a growing trend. We anticipate even greater use of grassroots campaigning on social media in future elections.

This type of analysis assists experts in understanding the new communication techniques and their utilisation across various information mediums.

Coverage dispersion per information medium for mention of Ivelin Mihaylov, the leader of Velichie.

Identrics 360:

The social echo—understanding the narrative resonance

When performing a situational report, analysing the content creation is just half of the process. In addition to analysing the “who” in a narrative, we are also interested in the “why.”

Authors, journalists, and everyday content creators sometimes push a story to the point where it becomes viral, and then information (bad) actors design and lead campaigns boosting engagement for their narratives. We are able to track this social echo of the online news medium to pinpoint early queues of organised campaigns, tracing the organic development of narratives and gauging the final impact of stories.

Identrics’ social echo tool goes beyond content analysis to reveal the “why” behind narrative spread. By tracking how news articles and social shares reverberate, we determine whether a story gains traction organically or through coordinated efforts.


The information coming from Identrics’ social echo can be utilised by disinformation analysts to monitor for coordinated or automated narrative engagement campaigns. Editorial teams also leverage it to keep track of their stories’ performance, as well as their colleagues’ engagement penetration.


For example, during the week preceding the election, on average we have detected little over 15 interactions per news article. This report can be further drilled down to show data per organisation and/or specific people, and even information sources and authors, to give analysts rich leads to follow.

An overall view of our dataset shows some interesting spikes in shared articles which lead to less than half engagement in reactions and comments, which might indicate coordinated campaigns in sharing articles.

Look who is writing—identifying content authors

Understanding who writes the stories—whether they are authored or anonymous—is crucial.

Content creators standing behind their work with their name, biography, and interests are key in keeping the integrity of the information flow. It is crucial to understand when narratives are woven by anonymous creators, which is why we analyse the authored vs. non-authored articles:

Authored and anonymous information artefacts across the whole infosphere.

Using Identrics tools, we identified networks of anonymous creators potentially driving propaganda efforts. The drill-down per parties and their leaders, as well per outlets, are utilised in tracking down networks of anonymous content creators which might indicate parts of organised information campaigns.

Analysing topic dispersion

Narratives form rapidly around elections, and analysing topic dispersion helps us gauge their depth and agenda.

Topic dispersion is vital in the methodology Identrics follows to maintain information quality by allowing for a quick jump-in point for analysis of a narrative’s informational depth, which might indicate its informational agenda.

Elections come, and it is always interesting to analyse the variety of topics covered by the different party leaders and the narratives they are prevalent in. This time, we examined key leaders like Atanas Zafirov (Bulgarian Socialist Party), Kostadin Kostadinov (Vuzrazhdane), and Kiril Petkov (WCC-DB) to see which narratives they dominated and why.

Topic dispersion for mentions of Bulgarian Socialist Party’s Atanas Zafirov.
Topic dispersion for mentions of the reformist WCC-DB’s Kiril Petkov.
Topic dispersion for mentions of the populist Vuzrazhdane’s Kostadin Kostadinov.

Democratising narrative analysis with predictive clustering

An addition to this sitrep is our predictive and unsupervised narrative clustering model.

Identrics’ predictive narrative clustering model allows analysts to explore organically formed narratives without relying on preconceived notions. This frees disinformation and misinformation analysts from designing searches based on possible preconceptions and allows them to pick a narrative to analyse based on its natural formation of people’s reactions to it.

This approach lets the data reveal the key narratives.

Combatting synthetic propaganda

With the rise of generative AI, synthetic content is increasingly used for propaganda, as part of cognitive warfare efforts. Coordinated communication attacks are a segue to state capture, and they have become easier to design.

Our WASPer models and taxonomies detect synthetic content, classify propaganda types, and help mitigate their impact.

By analysing whether a piece of content is synthetic, we assist communication professionals, editorial, and moderation teams in identifying and combatting propaganda. This is coupled with our narrative clustering to give another angle to inform narrative selection and analysis.

Hate speech detection

We have teamed up with editorial and moderator teams on media and social forums in the creation of the self-reinforcing hate speech detection model and tool for them to use as assistance in mitigating the polarised online speech and even more so now with the ease of access to generative AI technologies.

Media capture via comment sections targeted information attacks is a well-known tactic, unfolding in the Black Sea region.

Want to go deeper?

This sitrep is just the beginning. If you are interested in live insights on high-risk narrative analysis or a deeper dive into specific topics, we are here to assist.

We are excited to release Annex this December—a gateway to understanding and safeguarding information integrity at the national and corporate levels.

Ready for a deeper dive into the narratives? Let’s talk.