Google searches that start with a full question get fewer clicks anyway — so is AI really to blame for the drop?

This explores whether lower click rates on AI-summary searches might partly come from which searches get summaries, rather than only from the summaries themselves. The headline number is stark: in Pew's browsing data, people clicked a result link 8% of the time when an AI summary appeared, against 15% without one, and those sessions were more likely to end with no click at all Do AI summaries on Google reduce clicks to actual websites?. Your question asks the right thing. If summaries mostly appear on long, question-shaped searches, and those searches would have drawn fewer clicks anyway, then part of the gap belongs to the type of query, not to the AI. The corpus doesn't contain a study that separates these two effects, so it can't settle the question directly. It does give good reasons to think query shape matters.

The strongest lateral evidence comes from retrieval research, which treats the form of a question as a signal in its own right. One line of work finds that non-factoid questions (comparisons, debates, 'why' and experience questions) need different retrieval strategies from simple lookups. Some need to be broken into parts, and some need evidence gathered from several angles Does question type determine the right retrieval strategy?. A question like that rarely has one page that answers it, so the classic 'click the top link' behavior fits it poorly from the start. A related finding: lightweight surface features of a question, such as its length and structure, predict whether retrieval is needed about as well as expensive model-uncertainty methods Can question features alone predict when to retrieve?. If a question's shape tells a system how much searching it needs, it plausibly tells you something about how a person will search, too.

Research on deep research agents points the same way from the other side. Answer quality on complex questions keeps improving with more search steps, with diminishing returns, much as it does with more reasoning Does search budget scale like reasoning tokens for answer quality? Do search steps follow the same scaling rules as reasoning tokens?. Hard questions are answered across many sources rather than by one page. For a human, that could mean reformulating the query, giving up, or settling for a partial answer. None of those registers as a click on that particular search, so lower click rates for long questions could reflect people moving between searches rather than being satisfied.

One more twist cuts against treating a 'no click' as a sign the answer was good. In AI search arenas, users rated responses with more citations as more trustworthy, even when the extra citations were irrelevant Do users trust citations more when there are simply more of them?. Citations can work as a reassurance signal rather than an invitation to click. So even after separating out query type, an AI summary might cut clicks by making people feel informed instead of by actually informing them. The real test would compare clicks on matched queries of the same length and question type, with and without a summary. The corpus doesn't have that comparison, and anyone using the Pew figure as a pure causal effect should be asked for it.