A generation ago, a voter trying to make sense of a referendum question might have bought a newspaper, sat through a party political broadcast or asked a well-read friend. Today they are more likely to type the question into a chatbot and receive a tidy, confident answer before the kettle has boiled. Artificial intelligence is quickly becoming a new intermediary between citizens and political information, summarising election issues, explaining legislation and comparing competing arguments within seconds. The democratic potential is real, since information once buried in long policy documents can now be delivered in plain language almost instantly. Yet the convenience raises a harder question: if citizens increasingly rely on a small number of AI systems to understand politics, who decides what is surfaced, how it is framed and what is quietly left out?
Regulators have started asking the same thing. In September, the UK Competition and Markets Authority proposed requiring Google to give users greater choice over search services on Android and Chrome, including AI assistants such as ChatGPT and Perplexity. The regulator also proposed stronger attribution requirements so users can identify the sources behind search results. Behind the language of competition sits a democratic concern. If a handful of companies control the digital gateways through which citizens find political information, their design choices carry influence well beyond ordinary commercial rivalry.
The European Union has gone further. Since August 2026, transparency requirements under the EU AI Act have applied to certain AI systems, including obligations to identify AI-generated or manipulated material and to inform people when they are interacting with artificial intelligence. In September, European Commission officials also warned that AI and algorithm-driven information manipulation could distort public debate and weaken trust in democratic institutions and elections. These measures recognise that accessibility alone is not enough. Citizens also need to understand where political information comes from and whether it has been artificially generated.
The problem runs deeper because AI does not simply retrieve information; it frames it. An August 2026 political science experiment on AI-assisted information seeking found that biased information presented through AI significantly shifted participants’ attitudes towards ranked-choice voting, while highly agreeable, or “sycophantic”, responses produced a smaller but measurable effect. The findings do not prove that AI assistants determine political opinion. They do show how a conversational system that appears neutral can shape the context in which unfamiliar policies are first understood.
Restricting AI’s role too heavily, however, would ignore its democratic benefits. A citizen unfamiliar with economic policy, parliamentary procedure or a complex referendum question can use an AI assistant to get explanations that might otherwise demand considerable time or specialist knowledge. For politically disengaged citizens in particular, such accessible tools could lower barriers to participation rather than reinforce them.
The task, then, is to make AI a more trustworthy intermediary rather than to keep it out of political life. That means clear attribution so citizens can see where an answer comes from, genuine choice over which assistant they use, and independent scrutiny of how these systems handle contested questions. It also asks something of citizens themselves, who will need to treat a chatbot’s summary as a starting point for understanding rather than the final word. Newspapers, broadcasters and search engines have all shaped political debate before, and democracies learned, imperfectly, to hold them to account. AI deserves the same scrutiny, applied before its influence becomes too familiar to notice.
Zohaib Sadiq
Connect with us on socials
Website | Twitter (X) | Youtube | Linkedin | Tiktok | Facebook | Mailing List