Anne-Laure Le Cunff
A researcher in neuroscience at the Institute of Psychiatry, Psychology and Neuroscience at King’s College London, and author of “Tiny Experiments.”
More than 60% of Google searches in the United States now end without the user clicking a single link. We type a question, read an AI-generated summary, and leave with our answer. Google isn’t alone: Claude, ChatGPT, and upstarts like Perplexity do the same thing — take a question, return an answer, and compress what used to be a meandering journey through the internet into an instant arrival. The exploratory phase — clicking through links, stumbling onto unexpected pages, following a reference somewhere unplanned — is disappearing.
That may sound like good news for users. What could be wrong with a reliable answer, faster? Plenty. By shortening the distance between question and answer, these tools undermine curiosity and, paradoxically, threaten our ability to understand the world. I worked at Google about a decade ago, where we measured the value of content by engagement signals like clicks and scroll depth. The behavior Google rewarded — people exploring — is exactly what its AI products are now built to eliminate.
I left Google to study neuroscience, and the research explains the danger. Curiosity isn’t just a wish to learn a discrete fact; it’s a biological feature that requires a specific condition — a gap between what you want to know and what you find out. Researchers have shown that people in a state of curiosity, while awaiting an answer to an intriguing question, remember unrelated information they encounter far better than usual. Brain scans found that the wait activates reward circuits and primes the hippocampus to form new memories — a pattern seen in infants, children, and adults alike.
That window stays open only while the question remains unanswered. When AI answers in three seconds, the window shuts before curiosity can deepen. You got what you came for, but you lost what turns curiosity into learning: the adjacent article you might have read, the tangent you might have followed, the link between two ideas with no obvious connection. Researchers call this incidental learning, and it drives many serendipitous discoveries. When physicists Arno Penzias and Robert Wilson detected a persistent hiss in their radio antenna in 1964, they could have dismissed it as equipment noise; instead they kept asking what it was, and ended up discovering the radiation left over from the Big Bang.
The danger isn’t that people stop asking questions. It’s that questions become endpoints. The loss is trivial in any single case, but fewer detours and unexpected discoveries add up: over time, people trained this way get better at extracting ready-made conclusions than at building connections of their own.
No one is forced to use these tools — you can still wander and follow a chain of links into unfamiliar territory — but the default architecture of our platforms makes that less likely. AI companies that want to do right by their users have other design choices: keep sources visible, show competing explanations instead of compressing them into one smooth paragraph, and offer search modes that reward exploration over speed. The space between a question and an answer has value, and it shouldn’t be engineered away. If we build a world that delivers only what we ask for, we will lose the capacity to discover what we didn’t know to ask.
Source: Opinion essay, The New York Times, July 8, 2026