On 14 October 2026, AI developer Anthropic published an analysis of 250,000 conversations on chatbot Claude. It shows that 5.75 per cent of American conversations touch on politics, public policy and social issues.
Users rarely ask for voting advice, with only 0.01 per cent of questions explicitly about this. Around 0.51 per cent focus specifically on elections. More often, people ask questions to follow the news, understand public services or compare positions on social issues.
What this means for non-profits and government
People are using answer engines to answer societal questions and navigate public schemes. When your organisation publishes information about grants, provisions or social issues, AI models process those texts directly for the reader.
If someone asks an AI assistant about the conditions for a local energy allowance, the assistant reads council web pages and summarises the rules. If that information is unclear on the website, the assistant can miss or misinterpret details about the application process.
Practical steps for your communications team
Anthropic’s research comes from a one-week sample of American users. While exact figures vary across different periods and countries, the pattern is clear: citizens look for factual information using AI chatbots.
To make sure your information is handled correctly:
- Write policy details and scheme criteria in plain language without jargon.
- Structure information pages with clear subheadings and direct question-and-answer sections.
- Test popular AI assistants regularly to see how they explain your specific public services and social issues.
Source: Anthropic
