The click gap is large enough to change measurement
The difference between 8% and 15% is not a small ranking fluctuation. In this panel it represents a 7 percentage point gap. Relative to the no-summary result, the observed click rate was about 47% lower when an AI summary appeared.
That does not mean every lost click was caused by AI. Query type, intent, result composition, and summary eligibility differ. The practical conclusion is narrower: reporting that ignores the presence of an AI summary may mix two materially different search experiences.
Longer questions are more likely to produce summaries
Pew found an AI summary on 8% of searches containing one or two words, compared with 53% of searches containing ten words or more. Question-shaped and full-sentence searches also produced summaries more often.
This matters for intent analysis. Complex research queries are not distributed evenly across the results experience, so a brand can lose visibility on the questions that reveal the most context even when broad keyword rankings appear stable.
- Track query length and question form
- Separate summary and non-summary result experiences
- Measure cited presence and outbound clicks independently
Citation visibility and referral traffic are different signals
Most summaries in the study cited at least three sources, yet only 1% of visits with a summary produced a click on a cited source. A citation can still influence awareness or trust, but it should not be counted as a referral.
A useful AI search dashboard therefore needs at least three layers: answer presence, cited-source presence, and downstream site behavior. Combining them into one visibility score hides the commercial difference between being mentioned and being visited.