AI Thinks the Internet Is Reddit. We Have 8,616 Answers Proving It (growtika.com)

A recent investigation revealed a striking trend among OpenAI's GPT models, notably GPT-4.1, GPT-5.5, and GPT-5.6, where they frequently default to citing Reddit, even when prompts did not mention the platform. Out of 8,616 API runs, these models referenced Reddit in 998 instances, including 298 responses to prompts that never explicitly requested it. In contrast, the smaller GPT-4.1 mini model showed significantly less reliance on Reddit, indicating varying behaviors across different configurations. The models construct their own web-search queries, consistently incorporating Reddit, which raises questions about their inherent biases towards community-driven sources. This finding is significant for the AI/ML community as it underscores the models' tendency to prioritize Reddit as a primary source for community opinions, potentially shaping user interaction with AI tools and influencing the information landscape. The implications extend to brands and marketers, who may find that AI-generated responses are increasingly influenced by Reddit, regardless of user input. This highlights the need for businesses to adapt their strategies around this phenomenon, as the presence of community insights in AI-generated content becomes a norm, reshaping how users seek and receive information.