October 1, 2026
Page One Doubled. Clicks Didn't. I Sorted 4,944 Search Console Queries by Intent to Find Out Why.
Five local businesses, two 28-day windows, every query Search Console would give me, sorted into branded, local service, generic service, and informational. Page-one impressions on local service searches went up 129 percent. Page-one clicks went up 7 percent. Informational queries were 1 percent of the impressions to begin with. Half the clicks came with no query attached at all. The fight is not where the AI Mode headlines say it is.
AI Mode is eating informational search. Your reports can't see it happen.
The story going around this week has two halves. First, Google keeps widening the set of searches that open straight into an AI answer, mostly the informational ones: how long, how much, can you, what is. Second, nobody can measure it. AI Overview and AI Mode traffic gets folded into the same report as everything else, so you watch clicks slide and cannot point at the cause.
The second half is not a rumor. Google's own documentation says sites appearing in AI features like AI Overviews and AI Mode are included in the overall search traffic in Search Console, inside the Web search type. No separate line. And the big-site damage is real: HubSpot says organic traffic for its own customers fell 27 percent year over year.
Fine. But every one of those stories is about publishers and SaaS blogs that lived on explainer traffic. I work for pool builders and dentists. So I pulled the Search Console data for five of them and sorted every query by what the searcher wanted. If AI Mode is eating informational search, I wanted to see how much informational search a local business had to lose.
4,944 queries, four buckets, two windows
The setup. Five clients: a pool builder, a wrap shop, a window company, an HVAC company, and a dental practice. Two 28-day windows six weeks apart, July 18 to August 15 and August 31 to September 28. Search Console hands over up to 500 queries per business per window, which came to 2,444 queries in the first window and 2,500 in the second. Each query went into one bucket. Branded: it has the business name in it. Informational: it starts with how, what, can, should and friends, or asks about cost, price, types, or how long. Local service: near me or a place name. Generic service: everything else, the bare ac repair and pool builders.
Informational was never the money. 124 informational queries in the first window, 102 in the second. 396 impressions, then 642. That is about 1 percent of the impressions in the sample. Clicks: 2, then 4. Across five businesses. A window company ranked in the 80s for how long do replacement windows last and got exactly what the 80s get you: nothing. If AI Mode eats all of it, these five businesses lose four clicks a month.
Local service is where it hurts. Impressions on near-me and city searches went from 23,238 to 29,517. Rankings improved, average position from 22.9 to 16.3. Page-one impressions went from 7,484 to 17,117, up 129 percent. Page-one clicks went from 76 to 81. Up 7 percent. Click-through on page one fell from 1.02 percent to 0.47 percent. In the top three positions it got uglier: 16 clicks on 502 impressions in the first window, 4 clicks on 783 in the second. Most of the page-one gain came from two of the five, the dentist and the HVAC company, and the click-through drop held for both.
Branded is the bucket that still clicks. 57 branded queries in each window. Click-through went from 6.54 percent to 9.55 percent, and branded queries carried 56 percent of the sampled clicks in the first window and 67 percent in the second. Honest footnote: one client supplied most of that jump. Strip it out and branded clicks were 140 and 130. Even then, branded click-through runs about twenty times the local service rate in both windows.
Two more things fell out of the sort. Full-sentence searches, seven words or more or ending in a question mark, grew from 81 queries and 380 impressions to 94 queries and 656 impressions. My favorite: can i book a consultation for teeth whitening in [city]? The dentist ranked 2.2 for it. 46 impressions. Zero clicks. People are typing prompts into Google now, and Google is answering them. And the 500 queries Search Console gave me covered only 50 percent of the clicks in the first window and 52 percent in the second. The other half arrived with no query I could see.
The local fight is page one, not explainers. And half of it is dark.
For a local business, AI Mode taking the informational queries is a rounding error. The real squeeze is the near-me and city searches, where my clients climbed and the clicks did not come with them. I can't tell you from Search Console whether an AI Overview, the map pack, or an ad absorbed those clicks, because Google reports all of it in one bucket. That is the measurement gap in one sentence. I wrote about the hidden-click side of it in the hidden clicks piece and the zero-click side in the AI Overviews piece. This is the third leg: sort by intent and the decline stops looking random.
Branded holding up is the tell. A person who reads an AI answer that names you, then searches your name, shows up in the branded bucket, not the service bucket. If the answer names someone else, you never see that person at all. So the service bucket is where you lose and the branded bucket is where you find out whether the answers are naming you.
Ranking higher is not the fix. My clients ranked higher and got the same clicks. The fix is being the business the answer names, and being able to prove it with something better than a Search Console screenshot.
Triage by intent. Measure the answer, not just the rank.
One. Sort your own queries. Export the Search Console queries, mark branded, informational, local service, and generic service, and put the click-through for each bucket on one line per month. The total line goes in the trash. The four lines tell you where the money moved.
Two. Stop funding explainers for traffic. Write them to be cited. The wrap shop ranks around position 8 for can you wrap a car with dents, 117 impressions, 1 click. The click is gone. The citation is not. One tight, specific paragraph on the page beats a 2,000-word guide that nobody clicks to read.
Three. Put the effort where the money is. Local service pages, the Google Business Profile, reviews, and the sources AI answers pull from when someone asks who to call. Page one is crowded with answers now. The business named in the answer gets the call.
Four. Make branded search the scoreboard. If AI answers name you more often, branded queries and branded clicks go up. Track that line monthly next to the service line. Flat brand with rising service impressions means you are visible and unnamed.
Five. Measure AI answer presence somewhere Google does not grade its own homework. Search Console will not separate it, so the prompt-level number has to come from outside.
The number Search Console won't give you
The AI Visibility Toolkit tracks brand mentions, citations, and sentiment across ChatGPT, Perplexity, Gemini, Google AI Mode, and Claude on a steady weekly cadence. That is the missing line in my sort: for the near-me and city prompts where my clients climbed and the clicks stayed flat, it shows whether the AI answer named them, named a competitor, or cited a directory instead. Same prompts every week, so a drop means something.
The branded line needs feeding too, and branded demand gets built off Google as much as on it. The Social Toolkit is built for planning and scheduling the posts that keep a local business's name in front of people between searches, so when an AI answer mentions them, the name is already familiar enough to search.
Informational was a sideshow. Page one is the main event.
Five businesses and two windows is a sample, not a study, and seasonality is in there. But the shape was clear. The informational queries everyone is mourning were 1 percent of the impressions. Page one got twice as visible and earned the same clicks. Branded kept clicking. And half the clicks came without a query attached.
So stop arguing about whether AI Mode killed SEO. Sort your queries by intent, put your money on the searches that still buy something, and measure whether the answer names you. If you only watch the total, you are watching the wrong number.