A bachelor thesis built on Semantic Visions' svEye, tracking how the Novo Nordisk narrative evolved from diabetes treatment to systemic risk

When Ozempic became a global weight-loss phenomenon, the warning signs were already in the data – months, in some cases more than a year, before they reached mainstream headlines. A bachelor thesis built on Semantic Visions' svEye platform tracked exactly how the media narrative around Novo Nordisk's semaglutide drugs evolved from a niche diabetes treatment into a story of off-label demand, shortages, and systemic risk to vulnerable patients. Using real-world OSINT data drawn from global news media in twelve languages, the case study shows how AI-based media monitoring surfaces weak signals early enough to widen an organisation's decision-making window.

Semantic Visions provides this kind of insight to corporates, financial institutions and public bodies, turning millions of articles a day into structured early signals on risk, supply chains and reputation. The same data and tools are open to universities and students for research, and this article looks at one such project.

Key takeaways

• svEye flagged the "weight-loss drug" narrative from February 2021 – more than a year before it became a mainstream availability story in mid-2022.
• The South Africa insulin-access signal surfaced in svEye as a handful of articles around May–June 2024, roughly five months before it reached global outlets like The Guardian.
• Three connected risk narratives – reputational (emerging from early weight‑loss coverage in 2021), supply‑chain (shortage reports from 2022), and systemic access to medicines (South Africa in 2024) – tracked across six years.
• Signals were detected across twelve languages via a multilingual ontology, surfacing local-market risk regardless of source language.

A Student Case Study: Tracking Ozempic in Global Media

Barbora Lacková, a student at the Prague University of Prague University of Economics and Business (Vysoká škola ekonomická v Praze), chose Semantic Visions’ data as the backbone of her bachelor thesis, “Artificial Intelligence in Media Monitoring as a Tool for Information Risk Management.” Building on the svEye platform, she worked with real‑world OSINT data to track how media narratives around Novo Nordisk and its drugs Ozempic and Wegovy evolved over time. The case study follows Ozempic’s journey from a diabetes treatment to a highly publicised “weight‑loss drug,” and shows how this narrative shift contributed to concrete risks, including supply disruptions and wider reputational and systemic impacts. Throughout the thesis, svEye serves as the primary tool for identifying and analysing these signals in global media coverage.

Methods Behind the Case Study

To build her case study, Barbora worked directly in svEye, Semantic Visions’ OSINT and risk‑intelligence platform. She learned to navigate it largely on her own, as it is designed not just for professional analysts but also for non‑expert users, requiring only targeted explanations about how the underlying data and ontology are structured. For her Novo Nordisk analysis, she combined the platform’s two core modules: Insight Search, which offers high‑level overviews, timelines of key events and AI‑generated summaries, and Article Search, which lets an analyst drill down into individual articles and long‑term narrative development.

Within Insight Search, she used svEye’s Risk Intelligence Assistant (RIA) to ask questions in natural language, including how the Ozempic narrative changed after the SELECT trial, and to work with instant syntheses generated from the platform’s own data. Article Search then served as the analytical backbone for the graphs in her thesis, covering everything from the earliest mentions of Ozempic as a weight‑loss drug to reports of shortages and the South African insulin case. Alongside her hands‑on work in svEye, Barbora also conducted semi‑structured interviews with Semantic Visions experts to deepen her understanding of the platform, the data it provides, and the role of AI in media monitoring, which in turn helped her interpret the findings with greater confidence.

The Media Narratives in the Ozempic Story

As mentioned above, the case study follows the development of media narratives around Novo Nordisk’s semaglutide drugs. Barbora’s findings are particularly interesting, as they reveal three distinct but connected narratives that emerged over the last six years, with two of them first appearing in only a handful of articles before making wider headlines.

Let’s take a closer look at Barbora’s findings.

Narrative 1 – From Diabetes Treatment to “Aesthetic” Weight‑loss Drug

In the early phase, Ozempic appeared in the media primarily as a treatment for type 2 diabetes, but from February 2021 a few articles already described semaglutide as a prescription option for weight loss. These mentions stayed close to zero through 2021–2022, yet, as the graph shows, in svEye they were visible as an early warning that a new narrative was beginning to form. This information risk ultimately developed into a systemic issue. 

Later, this storyline was amplified on social media, especially TikTok, where celebrity use turned Ozempic into a trend. By May 2022, news coverage had begun to connect this TikTok wave with concerns about availability.

Narrative 2 – Off‑label Demand and Drug Shortages

The second narrative focuses on what happened once the weight‑loss storyline gained mainstream traction. Media reports began to describe unusually strong demand and shortages of Ozempic and Wegovy, explicitly linking these problems to off‑label use by people without diabetes, which resulted in the FDA adding both drugs to its shortage list.

In svEye’s timeline, articles about semaglutide supply issues appeared in 2022 and grew in volume, with a visible spike in February 2025 when the FDA announced the end of the shortage.

SELECT Study – A Turning Point in the Narrative

A key turning point in the overall storyline was the publication of the SELECT trial results in August 2023, showing that semaglutide can reduce the risk of cardiovascular disease by 20 percent for certain patients. Using svEye’s integrated AI assistant, Barbora explored how coverage shifted after SELECT: from a focus on “Hollywood shot” and “weight‑loss injection” towards semaglutide as a therapy for obesity with proven cardiovascular benefits. 

The corresponding graph shows a sharp rise in articles that frame semaglutide in connection with heart‑disease prevention, illustrating how a single scientific milestone can reframe an established media narrative.

Narrative 3 – South Africa and Systemic Risk

The third narrative zooms in on a geographically specific but highly sensitive episode in South Africa. In May 2024, Médecins Sans Frontières reported a shortage of insulin pens; in June, the National Department of Health stated that Novo Nordisk was prioritising the more profitable Ozempic line over insulin, which threatened patients dependent on state‑funded treatment. This narrative demonstrates how an information risk can become systemic, affecting access to essential medicines in a lower‑income market.

As the graph shows, this geographically distant event appeared in svEye as only a handful of articles — something traditional manual monitoring could easily overlook, only catching it once the story had spread to some of the global media outlets such as The Guardian in October 2024.

What svEye Reveals: Weak Signals and Context

Barbora uses the Novo Nordisk case to show not just what happened in the media, but how AI‑based monitoring helps organisations see these shifts early and interpret them responsibly.

Seeing Weak Signals before They Become Headlines

Across all three narratives, a recurring theme is weak signals: small clusters of articles that hint at a new storyline long before it becomes mainstream. In the “aesthetic weight‑loss drug” narrative, the number of mentions is almost zero for a long time, yet they are still detectable as a pattern when you aggregate global coverage by topic and entity. In the South Africa narrative, only a few pieces mention insulin shortages and prioritisation of Ozempic, but together they already point to a potential systemic risk. In Barbora’s framing, AI helps surface these faint patterns across languages and regions so that organisations do not have to wait for a full‑blown media crisis before they react.

Ontology As a Semantic Backbone

Another element Barbora highlights is svEye’s ontology – the structured network of entities and topics that sits underneath the AI models. Instead of treating each article as an isolated text, svEye uses a large, multilingual ontology and a knowledge graph to connect companies, locations, commodities and events into structured real‑world events. It also supports cross‑lingual analysis: the same concept is recognised whether it appears in English, Chinese or Spanish, which is crucial when early signals emerge first in local media.

Human‑in‑the‑loop, Not Automation on Autopilot

Despite the heavy use of AI, the thesis is very clear that svEye is not an automated 'truth machine'. Models help with language understanding, entity recognition, clustering of events, and summarisation, but human expertise remains central at two levels. First, domain experts and linguists help build and refine the ontology and models, correcting errors and tuning how topics are detected. Second, analysts using svEye interpret the results: they decide which weak signals matter, how to read conflicting sources, and what the business or public‑health implications are. Barbora’s findings support the idea that the most effective setup is a partnership: AI does the heavy lifting on volume and speed, while people provide context, judgement and ethical oversight.

From Commercial Risk to Academic Insight

The case study shows that AI-based media monitoring can genuinely function as a tool for information risk management, not just as a way to count mentions. By tracking the three Ozempic narratives, svEye surfaces weak signals about new storylines, shortages, and systemic issues before they fully break into mainstream global coverage, which widens the decision-making window for any organisation watching them. In the thesis conclusion, this is summed up as AI helping in three main ways: early detection of weak signals, reduction of information noise, and the use of semantic ontology across languages and regions.

While these features are highly useful for commercial risk management, the project also highlights a major benefit for academia. The exact same data and analytical tools that support these corporate use cases are available to universities and students for their own research.

Barbora's thesis is a clear example of how a student can take real-world OSINT data, learn the system quickly, and build a case study largely on their own. By making these commercial tools available to universities, students get the chance to use global news data and advanced analytical tools for their own research projects instead of just relying on theoretical examples. Ultimately, it shows that when students are given access to the same platforms used by professional analysts, they can independently explore complex global topics and build real-world skills in interpreting how AI tracks information risk.

You can find the whole thesis by Barbora Lacková here.

And now, the long game

Layered on top of all this, a newer narrative is starting to surface — side effects. Because semaglutide has now been in widespread, sustained use for years rather than months, longer-term effects are only now becoming visible: cases of NAION-related sudden vision loss serious enough to reach the drug's label, a growing body of gastroparesis litigation, and early signals around muscle and lean-mass loss. It's the same weak-signal dynamic the whole Ozempic story has followed — a handful of reports today that the data lets you see forming well before they become tomorrow's headline.