Internet users increasingly rely on search engines to access information, yet the nature of these platforms is shifting. Synthesised AI responses now occupy the space where clickable links once appeared, pushing original sources further down the page. Simultaneously, growing numbers of people bypass search engines altogether, turning instead to chatbots as their first port of call for answers.

This transformation introduces a critical opacity problem. When AI systems generate responses, the mechanisms behind those answers—which sources informed them, how reliable those sources are, and what reasoning shaped the output—often remain hidden from users.

The consequences ripple across multiple dimensions. Users encounter supposedly complete answers without incentive to consult underlying sources, undermining the habit of independent verification. Simultaneously, the shift threatens the financial foundations of journalism and publishing. As audiences visit media websites less often, advertising and subscription revenue dry up. Yet this creates a paradox: AI systems depend on the very media outlets and publishers whose business models they are eroding to source trustworthy content for their own responses. Over time, this dynamic risks degrading the quality and availability of reliable information—a cornerstone of democratic participation.

Since 2023, AlgorithmWatch has conducted regional election case studies to assess how accurately AI systems handle political questions and what dangers this poses to democratic deliberation. The organisation's research has covered three separate investigations into Microsoft Copilot (previously known as Bing Chat), Google Gemini, models developed by OpenAI, and Google's AI Overviews.