AI Information Map: Which Sources AI Systems Surface in Hungary

The Media and Journalism Research Center (MJRC) has published the Hungary edition of its AI Information Map, a research project examining which sources AI systems make visible when people ask questions about news, politics and public affairs.

The new Hungary study tests 10 major AI systems, including ChatGPT, Gemini, Claude, Perplexity, Copilot, Grok and DeepSeek, using the same questions in English and Hungarian. The research draws on around 320,000 archived answers and tracks not only which sources the systems use, but also which they recommend, caution against or associate with disinformation.

The findings show just how differently AI systems can construct the same national information environment. EU institutions, Reuters and Wikipedia dominate the overall sourcing landscape, while the choice of language dramatically changes which journalism becomes visible. Hungarian newsrooms account for around one fifth of citations when questions are asked in Hungarian, but less than 4% when the same questions are asked in English.

Hungary is the third country published under MJRC’s AI Information Map, following earlier studies of Romania and Spain. The wider project investigates how AI is becoming a new intermediary between audiences and journalism: selecting, ranking, recommending and sometimes overlooking sources before users ever reach the original reporting.

The first phase covers 10 countries. Research on Brazil, Germany, India, the Philippines, Serbia, South Africa and the United Kingdom will follow, with further countries and language environments to be added as the database expands.

The aim is to build a comparative picture of a question that is becoming increasingly important for media pluralism: when AI answers the news, whose journalism gets seen?