When an AI gives you the answer upfront, do you still click through to where it came from?

This explores whether AI summaries act as a doorway to sources or as an endpoint. The short answer from the corpus: mostly an endpoint. Pew's browsing-panel analysis of nearly 69,000 real Google searches found people clicked a search result 8% of the time when an AI summary appeared, versus 15% when there was none. Sessions were also 10 percentage points more likely to end with no click at all Do AI summaries on Google reduce clicks to actual websites?. Industry data points the same way. Ahrefs found the top organic result loses about 58% of its clickthrough rate on queries that show an AI Overview How much are AI Overviews actually reducing organic search clicks?. A natural experiment comparing Wikipedia's language editions found English Wikipedia lost roughly 5% of its search referrals once AI Overviews launched in English, with German and French serving as controls Does AI search summaries divert traffic away from Wikipedia?. One caveat: these notes measure clicks on ordinary search results. None of them isolates clicks on the citation links inside the summary itself, so the corpus can't say how often those get used.

Eye-tracking work shows why. The summary takes over the spot on the page where people look first. In an RMIT and Microsoft study, searchers fixated on AI Overviews far longer, and the share of attention going to the first-ranked result fell from 31% to 9%. Trust stayed equally high for both Where do searchers look when AI Overviews appear?. People aren't skipping the links because they distrust them. Their attention is already spent by the time they'd reach them.

The less obvious finding: better summaries make this worse. In Nextdoor experiments, LLM-written notification summaries were measurably more informative and got fewer clicks, because a good summary removes the reason to open anything Does better summary writing actually increase user engagement?. Summary quality and click-through can pull in opposite directions. The e-commerce work on ReLSum shows the reverse can be designed for. When summaries are trained against what users actually do next, rather than for fluent prose, engagement goes up Can reinforcement learning align summarization with ranking goals?. So whether a summary ends the session depends partly on what it was built to optimize.

Two threads complicate the simple story that AI answers replace the web. A cross-surface panel study found people more often open AI assistants after searching and browsing than before, by about 20 percentage points. That suggests many people still go to the web first and use assistants to finish the job Do people use AI assistants before or after searching?. Skipping the click may also have a learning cost. Across seven experiments with more than 10,000 people, those who learned from ChatGPT summaries reported learning less and felt less ownership of what they knew. The advice they then wrote was rated sparser than advice from people who had searched the web Does learning from AI summaries produce shallower knowledge than web search?. So the ended session isn't only a traffic problem for publishers. It may also mean shallower understanding for the reader who stopped at the summary.