Google Search Is Answering More and Linking Less. Here's Why Your Phone Can Still Ring More.

You have probably seen a headline in the last week claiming Google has replaced search with AI, and that every query now returns an AI answer instead of a list of links. A few of those headlines are ahead of the facts. But the underlying trend is real, it is measurable, and it changes how we run campaigns for home service businesses.

Here is what actually happened, what the research shows, and why we think the next two years favor contractors who invest rather than pull back.

First, What Actually Changed

At I/O in May 2026, Google announced that Gemini 3.5 Flash became the default model inside AI Mode, and that AI Mode had passed one billion monthly users with queries more than doubling every quarter. In the same announcement, Google also said this: "You'll continue to get a range of results from Search, just like you do today." (Google, May 19, 2026)

That is a model upgrade within a feature. It is not the same thing as Google deleting the blue links. On June 8, Google Search VP Rajan Patel said plainly that Google does "not plan to make AI mode the default behavior for Chrome search."

We want to be straight with you about this, because a lot of marketing agencies are currently using the scarier version of the story to sell panic. The classic results page is still there. The Web tab is still there.

What is true, and what matters far more, is this: AI Overviews now reach more than 2.5 billion monthly users, they appear on a growing share of commercial searches, and they are measurably reducing how often people click through to websites. That trend is real and it is accelerating. You do not need the exaggerated headline to justify changing your strategy.

The Click Decline Is Real

Three independent sources, using different methods, point the same direction.

Pew Research Center tracked the actual browsing behavior of 900 U.S. adults across 68,879 searches in March 2025. When an AI summary appeared, users clicked a traditional search result on 8% of visits. When no AI summary appeared, they clicked on 15%. Users also ended their browsing session entirely 26% of the time with an AI summary present, versus 16% without. (Pew, July 2025)

Worth noting: Google publicly disputed Pew's methodology. We are including it anyway because it is one of the only non-commercial, panel-based studies on this question, and because the direction matches everything else.

SparkToro, using Similarweb panel data, found that 68.01% of Google searches ended without a click between January and April 2026, up from 60.45% in 2024 and 49% in 2019. That is a 7.5 point jump in 24 months, the fastest move in a decade. (SparkToro, June 2026)

The same study found AI Mode accounted for only 0.34% of all Google searches in that window. A lot of people have tried it. Very few use it for most queries yet. The click decline is being driven by AI Overviews on the regular results page, not by AI Mode.

Ahrefs analyzed 300,000 keywords and found that the presence of an AI Overview correlates with a 58% lower click-through rate for the top-ranking page. Position two saw a 50.8% drop, position three 46.4%. (Ahrefs, February 2026) Ahrefs is careful to say "correlates," not "causes," and so are we.

Google's own position is that total organic click volume has been "relatively stable" year over year, and that it is sending "slightly more quality clicks" to websites, where quality means users who do not immediately click back. (Google, August 2025) That post contains no numbers. Take it for what it is.

Local Search Is a Different Animal

This is the part most coverage skips, and it is the part that matters most if you sell tree work, HVAC, roofing, or plumbing in one metro.

Almost all of the alarming click data above is measuring informational searches. "How do I know if my tree is dying." "What causes a furnace to short cycle." Those queries are getting absorbed by AI, and they were never the ones that made your phone ring anyway.

Transactional local searches behave differently. Whitespark studied 540 local queries across three cities and six verticals, including plumbers. Local-intent queries returned an AI Overview 15% of the time and a local pack 93% of the time. Informational queries were the mirror image: 92% AI Overview, 6% local pack. (Whitespark, May 2025) That data is over a year old now and the gap has narrowed since, but the structural split holds.

Sterling Sky's State of Local SEO 2026, published in June, is the freshest credible local dataset we have found. Two findings stand out:

AI local packs appeared on only about 7% of tracked keywords. The traditional three-pack is still doing the heavy lifting for local intent.

But where AI local packs do appear, they are brutal. Across 322 markets, 88% showed fewer unique businesses than the regular pack. AI packs surfaced 5,943 unique businesses where regular three-packs surfaced 18,330, roughly a third as many. They often show one or two businesses instead of three, and they do not include call buttons.

Fewer winners, more screen time each. If you are consistently in the top two locally, this is good news for you. If you are hovering at four through seven, this is the threat.

People Still Click, and They Still Read Reviews

Sagapixel ran a behavioral study on transactional AI Mode sessions, observing 52 participants for about 22 hours on high-involvement service queries. The results run against the doom framing:

• 69% of AI Mode sessions still resulted in a click through to a business website
• Only 27% felt ready to make a decision from the AI summary alone
• 89% clicked multiple businesses, averaging 3.7 per session
• 74% read reviews before deciding

(Search Engine Land, December 2025)

The sample is small and the vertical is healthcare rather than the trades, so treat it as directional. But the logic transfers cleanly. Nobody lets an AI pick who removes the eighty foot oak leaning over their house. When the job is expensive, on their property, and hard to reverse, people verify before they call.

Why Fewer Clicks Does Not Mean Less Money

Here is the part we want owners to actually internalize.

Traffic is an input. Booked jobs are the output. Those two numbers have never been as tightly coupled as reporting dashboards imply, and AI search is pulling them further apart in a way that mostly helps well-run businesses.

Think about what the AI is actually doing before someone reaches you. It reads your service pages. It reads your Google Business Profile. It reads your reviews. Then it hands the homeowner a short list. The people who used to visit six websites, skim, and bounce from five of them are now getting filtered out upstream. They never show up in your analytics. Your session count drops.

What is left is a smaller number of people who have already been pre-qualified by your own content, summarized through the AI, and who are now calling you.

There is early data supporting this. Invoca's 2026 Lead Conversion Benchmarks, built on more than 70 million calls across ten industries and published July 13, found that calls referred from ChatGPT became qualified sales leads 49% of the time, against a 39% all-channel average and 43% for Google Business Profile.

Note the honest caveat, because it matters: those AI-referred leads closed at 40%, slightly below the 42% all-channel average. So the accurate version is that AI-referred callers arrive better qualified, not that they buy more reliably once they are on the phone. Anyone telling you AI traffic converts twenty times better is quoting a number with no primary source behind it.

On the ecommerce side, Adobe's Q2 2026 AI Traffic Report found AI-referred traffic converted 42% better than non-AI traffic in March 2026, a full reversal from 38% worse a year earlier, with 48% more time on site and 37% higher revenue per visit. (Adobe, April 2026) That is retail, not lead generation, so we would not port the exact figure to a tree service. The direction is still worth knowing.

The math that matters for you is simple. If your site gets 40% fewer visits but the visitors who arrive convert at two or three times the old rate, you booked more work with a worse-looking traffic graph. We would take that trade every time, and so would you if the dashboard were not shouting at you.

Where the Leverage Actually Moved

This is what we are doing differently for clients based on the research.

1. Google Business Profile is now the single highest-leverage asset

This is the most important finding in this entire post. Profound analyzed more than 32 million AI Mode citation instances and found that google.com is now the second most-cited domain in AI Mode. Citations to Google's own pages grew 8.4x between April 15 and June 30, 2026, and that growth came almost entirely from Google Business Profiles and Product Knowledge Panels. The strongest movement was in local-intent categories, with home services named specifically. (Search Engine Land, July 2026)

Read that again. When Google's AI answers a question about local contractors, it is increasingly citing Google Business Profiles as the source. Your GBP is no longer a directory listing that sits next to your marketing. It is a primary input to the answer itself.

Every field filled. Every service listed with its own description. Current hours, service area, and payment methods. Fresh geotagged photos every month. This was always on our checklist. It is now the first line item, ahead of the website.

2. Review volume became a visibility input, not just a trust signal

A May 2026 study of 1,120 queries across seven verticals and thirteen cities found that plumbers appearing in AI Mode had a median of 546 reviews. Plumbers that did not appear had a median of 125. That is 337% more reviews for the businesses the AI chose to surface. (Local SEO Data, June 2026)

Reviews used to be the thing that convinced a homeowner to call you once they found you. Now they are part of how you get found at all. If you have 60 reviews and the company across town has 500, you are not just less persuasive, you may be absent from the answer.

The same study found that 28.5% of businesses in the traditional local pack were entirely missing from AI Mode for the same query, and that 999 businesses appeared only in AI Mode. It is a separate system with separate inputs, not a reskinned map pack.

3. Ranking in AI Overviews does not get you into AI Mode

Ahrefs compared 540,000 query pairs and found the citation overlap between AI Mode and AI Overviews is only 13.7%, even though the two systems produce answers that are 86% semantically similar. (Ahrefs, December 2025)

Same answers, different sources. There is no single optimization that wins both. The practical response is breadth: more pages, covering more specific questions, so you have more surface area for either system to find you.

4. Page one is no longer the only place citations come from

Ahrefs analyzed 863,000 search results and 4 million AI Overview citation URLs and found only 37.9% of citations come from the top ten organic results. Another 31.2% come from positions 11 through 100, and 31.0% come from beyond position 100. (Ahrefs, March 2026)

This is genuinely good news for contractors, and it is the strongest argument we have for writing deep service and location pages. A page that would never crack page one for a competitive term can still get quoted in an AI answer if it answers a specific question well. Content that used to be "not worth writing because we'll never rank" is now worth writing.

5. How people phrase the query changes whether you show up at all

The same 1,120-query study found that searching "divorce lawyer" returned zero business listings in AI Mode across ten cities, while "divorce lawyer near me" returned listings in 100% of cases, 38 out of 38.

Conversational, local phrasing triggers local results. Bare service terms often do not. This is why we write pages around how homeowners actually talk, not around keyword tools. "Emergency tree removal near me after a storm" is a real query. "Tree removal services" is a spreadsheet entry.

6. Ads are showing inside AI Mode, and home services is in the expensive bucket

SE Ranking tested 50,032 keywords across 20 niches at the end of June 2026 and found 29.45% of ad-eligible commercial queries returned a text ad in AI Mode. That rate scales sharply with cost per click: 24.33% for keywords under $2, 32.45% for $2 to $10, and 53.56% for keywords above $10. Home service terms sit squarely in that top bucket. (SE Ranking, July 2026)

Two things follow. First, paid is holding its position while organic clicks compress, which is exactly why we keep clients running ads even when their SEO is strong. Second, only 11.53% of advertiser domains also appeared as a cited source in the same answer. Paying does not buy you a citation. They are separate games and you want to be in both.

Be aware of the reporting limitation too. Google states you cannot opt out of serving ads in AI Overviews, and it does not offer segmented reporting for them. (Google Ads Help) If someone shows you an "AI Mode cost per lead," they made it up. The data does not exist.

7. Your phone is now part of your ranking

Google Search rolled out an AI feature that calls local businesses on the user's behalf. A homeowner searches for a service, taps to have AI check pricing, and Google's AI phones several businesses to collect availability and pricing, then reports back. It is live for U.S. Search users, and every call opens by announcing it is an automated system calling from Google. (TechCrunch, July 2025)

For tree services and HVAC companies, a growing share of inbound calls may be an AI agent price-shopping you against three competitors at once. Nobody has published volume data yet. But if your voicemail is full or nobody picks up between 11am and 1pm, you are now losing placements, not just calls.

What We Would Do This Quarter

If you do nothing else, do these five things in this order.

1. Audit your Google Business Profile field by field. Every service, every description, current hours, service area, photos from this month. It is the most-cited source in your category right now.
2. Build a review system, not a review campaign. The gap between 125 and 546 reviews is not closed with a one-time push. It is closed with a request going out after every completed job, every week, permanently.
3. Write the pages you skipped because you thought you could not rank. A third of AI citations come from beyond position 100. Specific questions, specific towns, real answers, full length.
4. Fix your phone coverage. Answer rate is now a visibility factor, not just a sales factor. Someone answers during business hours, voicemail is cleared, and the person answering can quote a range without transferring the call.
5. Keep the ads running. Paid placement is holding while organic clicks compress. This is the wrong quarter to cut the budget because a traffic graph looked bad.

Our Opinion

This section is our read, not established data.

We think reported metrics are going to get worse for nearly everyone over the next year, including for campaigns that are working well. Impressions will fall because AI answers show fewer businesses. Sessions will fall because the AI handles the research phase. If you judge your marketing by traffic charts, the next twelve months will feel discouraging even in a good year.

We also think the businesses with genuinely deep, specific, honest content are about to pull further ahead than they ever could under the old system. When the AI is doing the reading on the homeowner's behalf, the company with real pricing ranges, real explanations of how the work is done, and real answers to the awkward questions gets recommended. The company with four thin service pages gets skipped, and it does not even get the consolation of a bounced visit to show for it.

This is the part we want to be careful about, because it cuts against our own interest in a comfortable story: we do not think this is automatically good for every business. Consolidation to one or two names per query means the middle of the pack gets squeezed hard. If you are fourth locally, the old map pack gave you a share of the clicks. An AI answer that names two companies gives you nothing.

So the honest summary is not "fewer clicks, more money." It is "fewer clicks, and the money concentrates." Whether that helps or hurts you depends entirely on whether you do the work to be one of the names that gets returned.

The work has not changed much. Complete profile, steady reviews, deep content, a phone that gets answered. What changed is the penalty for skipping any of it.

Sources

Every statistic in this post links to its original source. Where a study had methodological limits, a small sample, or a commercial interest behind it, we said so in the text rather than in a footnote. If you want to check any of it yourself, the links are live.