As more people rely on AI to make everyday decisions, one question has become increasingly important: Where do those answers actually come from? According to a recent study by Muck Rack’s Generative Pulse team, AI models rely on different sources to build their responses, and the type of question someone asks can dramatically change the information an AI model chooses to surface.

How AI Models Retrieve Information

The study analyzed more than 25 million citations generated by ChatGPT, Gemini, and Claude across thousands of real-world prompts. The analysis found that AI models draw from a mix of sources, including Wikipedia, news outlets, reference sites, scientific research, and academic literature. For example, when asked about the best pickup truck, ChatGPT recommended the Ram 1500 and cited Cars.com, while Claude chose the Ford F-150 without citing a source.

The research also found that AI changes its sourcing strategy depending on what people ask. If someone asks about current events, such as egg prices or the housing market, AI is more likely to rely on recent news coverage. If they ask about product reviews or how-to questions, it’s more likely to draw from product reviews, reference sources, and corporate sites.

The Role of Journalism in AI Answers

Despite some concern that AI is replacing journalism, the study found that news reporting continues to play a significant role in shaping AI answers. More than one quarter of all citations came from journalistic sources, with over half of those articles having been published within the previous year. This highlights the importance of reliable and trustworthy sources in informing AI models.

As AI becomes a more common way to discover information, understanding where an answer came from may become just as important as the answer itself. By recognizing the sources that inform AI models, we can better evaluate the accuracy and reliability of the information they provide.

Original reporting: KRDO (Colorado Springs metro) — read the source article.