Yes, on the searches that show one. Google users clicked a regular search result about half as often when an AI summary sat at the top of the page. The part that matters for a trade business is that the searches which bring you actual work are the least likely to show a summary at all.
What the click numbers actually say
The Pew Research Center published this on July 22, 2025. Read the report itself rather than the many blog posts that repeated it.
Pew used the browsing data of 900 US adults who agreed to install an app that tracks the pages they visit. It took every URL they opened on a tracked device from March 1 to 31, 2025. That left 68,879 unique Google searches. Pew then re-ran those searches through a scraping service between April 7 and 17, 2025 to see which ones returned a summary. Some 18% of them did.
Then the split that everyone quotes. On visits where an AI summary appeared, people clicked a regular search result in 8% of visits. On visits with no summary, they clicked in 15% of visits. Nearly twice as often.
The links inside the summary did not make up the gap. Pew put those clicks at 1% of visits to a page that carried a summary.
People also stopped browsing sooner. 26% of visits to a results page with a summary ended the session there, against 16% of visits to a page without one.
The searches that pay you show fewer summaries
Whitespark, a local search firm, ran 540 queries across Houston, Phoenix and Denver in six kinds of local business, plumbers among them. It published the counts on May 12, 2025.
Sorted by what the searcher wanted, the gap is wide:
- Local intent, the “plumber Houston” kind: an AI Overview appeared for 15% of those queries.
- Informational, the “why is my water heater leaking” kind: 92%.
- A mix of the two: 97%.
The map held its ground on the local ones. Whitespark counted a local pack on 93% of local-intent queries, and on only 6% of informational ones.
Pew found the same shape from the other side. 8% of one or two word searches produced a summary, against 53% of searches of ten words or more. 60% of searches that started with question words such as who, what, when or why got one.
Put those together and the picture is plain. A short “electrician near me” search mostly still returns a map and a list of businesses. A long typed-out question mostly gets an answer at the top of the page.
So the exposure is not spread evenly across your site. Your advice pages sit in the part of search where summaries are normal. Your service and town pages sit in the part where they are not.
Google says clicks are steady and publishes no numbers
On August 6, 2025, Liz Reid, VP and Head of Google Search, wrote that “Overall, total organic click volume from Google Search to websites has been relatively stable year-over-year.” She added that average click quality has gone up, and counted a quality click as one where the visitor does not quickly click back.
That post contains no figure, no chart and no method. There is nothing in it you can go and check.
Say that plainly, and note that it cuts both ways. Pew shows its sample size, its month, its query count and its limits. Google offers an assurance. Neither of them measured your website.
What none of this proves
Pew compared two piles of searches side by side. It did not follow one website’s traffic before and after a summary turned up. The searches that trigger a summary are different searches, longer and more question shaped, so part of that gap may sit in the questions people typed rather than in the summary. Pew did not claim a cause, and nobody selling you anything should claim one either.
The sample is 900 people who volunteered to install a tracker, over one single month. Pew states that it covered Google only, since it could not reliably spot AI summaries on other search engines. So none of it tells you anything about Bing, ChatGPT or Perplexity.
Whitespark’s figures are 540 queries, three cities and six business types. The page does not say when the queries were run, and the people running them were not sitting in any of those three cities. Your town is not in it and your trade may not be either. Results pages get rebuilt often, and a count from last year is a starting point, not a fact about today.
Here is the part a vendor will skip. If nearly all your calls already come from the map and from people typing your town and your trade, a package sold on the back of a frightening click statistic is not aimed at your actual problem. Ask what it changes before you pay for anything, and be suspicious of anyone who promises a mention in an AI answer. No one controls that.
If your first instinct is to pull your site out of AI answers altogether, that is a real option with real costs, and it is worth reading what opting out actually does before you touch a setting.
What to check in your own numbers
None of the studies above is your business. Your own analytics is, and it is free.
Split your pages into two groups. Pages that answer a question, and pages that sell a job in a place. Compare the last twelve months for each group on its own, not for the site as a whole, since a site-wide total hides the thing you are looking for. A fall concentrated on the advice pages, while the service and town pages hold steady, matches the pattern in the research above. Check the calls and forms too, not only the visits, which is a separate check with its own method. Fewer visits and the same number of booked jobs is a different problem from fewer of both.
Telling an AI-driven visit apart from any other visit is a separate job with its own setup, and there is a piece on measuring traffic from AI answers that covers it.
Do not gut your advice pages after one bad month. Those pages are often what gets a business quoted in an answer in the first place, and search traffic moves around for a dozen reasons that have nothing to do with AI, a new site launch among them.
One thing is worth doing whatever the numbers say. Put the phone number, the service area and the hours in plain text on every page, so a person who lands once can act without hunting.
If you want to see what the assistants currently say about your trade in your town, a free scan asks them the questions your customers ask and writes down which businesses get named. It is a check, not a fix, and a person reviews it before it is sent. The ongoing search checks are a separate thing, and it is fine to look at the scan first and decide nothing.