The summary answered it. Nobody scrolled.
Every number published for this is somebody’s sample of somebody’s queries. The mechanism is general; the size is yours, and it is measurable on your own set.
You have probably read that AI summaries cut click through rates by some specific percentage. Treat every one of those figures as what it is: a measurement of a particular set of queries, in a particular category, at a particular moment, usually published by somebody selling a fix.
The mechanism generalises. The magnitude does not, and it is the magnitude people quote.
The mechanism, which does generalise
A composed summary sits above the results and answers the question. Readers who wanted only the answer stop there. Readers who wanted to act — buy, compare, book, contact — carry on. So the loss is not spread evenly across your queries; it is concentrated in the ones a paragraph can finish.
- Definitional questions lose the most. "What is an issue tracker" has no reason to produce a click once it is answered.
- Single-fact questions lose almost everything. Opening hours, prices, dimensions, dates.
- Comparisons hold up. A paragraph can list options; it cannot resolve a trade-off for a specific buyer.
- Transactional queries hold up best, because the answer is not the destination.
Measuring it on your own queries
This is a comparison you can run without anyone’s benchmark. Segment by whether a summary appears, hold position constant, and compare click through within each segment over the same window. If the gap is large in the summary segment and flat in the other, you have measured your own number instead of borrowing one.
What replaces the click
A mention. When the summary names you or cites your page, your brand enters the reader’s decision without a session ever appearing in analytics. That is the surface worth holding, and it is the reason a report that only counts sessions now under-reports what is happening to you in both directions.
The short answers
How much do AI Overviews reduce click through rate?
There is no honest general figure, and we will not quote one we did not measure. Every published number is a sample of particular queries in a particular category. The mechanism generalises; the size is specific to your query set and is measurable on it.
Which queries lose the most clicks to AI summaries?
Definitional and single-fact questions, because a paragraph finishes them. Comparisons and transactional queries hold up, because the answer is not the destination.
How can I measure the impact on my own site?
Segment your queries by whether a summary appears, hold position constant, and compare click through within each segment over the same window. A large gap in the summary segment and a flat one elsewhere is your own number.
Does being cited in the summary help?
It is the replacement for the click. Your name enters the reader’s decision without a session appearing in analytics, which is also why session-only reporting now under-reports what is happening to you.
Will this get worse?
Unknown, and anyone forecasting it precisely is guessing. What is knowable today is which of your questions trigger a summary and whether it cites you, and both are measurable now.
Find out in under ten minutes
One free check, no account and no card. You get the answers the assistants gave, not a score.
Every figure quoted here comes from a real question put to a real assistant, saved with the date and the assistant that answered it. This piece was written to answer the search "ai overviews traffic impact".