After AI Search, Source Diversity Becomes Human Work

Traditional search places multiple results in front of us and leaves the comparison to the user. Generative AI compresses that stage into a coherent answer. Reading becomes easier, but the cognitive task changes: we can obtain “an answer” more quickly without necessarily seeing what other sources and paths sit behind it.

Microsoft researchers analysed more than 200,000 real-world human–ChatGPT interactions. They found that almost 80 per cent of the queries were not readily answerable through traditional search, suggesting that generative AI expands the space of questions people can explore. Yet for queries that could be compared with Google search, AI responses showed lower information diversity across most topics. The study also found that the diversity of AI outputs influenced how broadly users continued to explore.

So after AI search, the human task is not only to decide whether an answer is correct. It is also to reopen the source space deliberately. For important questions, checking different sources, viewpoints and original materials is not a leftover habit from older search. It is a new responsibility in a generative-search environment. AI can compress information for us, but it should not decide which voices remain visible.

From Searchable to Non-Searchable: Generative AI and Information Diversity in Online Information Seeking