HHVAC Foundry

geo ai

Google AI Mode vs Traditional Search for Home Services

Published 2026-08-13Updated 2026-08-1312 min read7 verified sources

Google's AI Mode answers a home-services question with one AI response plus linked sources, where traditional search returns ranked pages and a map pack. The measured differences: 100+ million monthly users, citations matching AI Overviews' only 13.7% of the time and Google's top-10 URLs roughly a third of the time, and a local pack in 13.49% of queries.

This comparison covers AI Mode specifically — how it selects sources, what its citations do differently, and what the early data says for local service businesses like HVAC companies. AI Overviews, the automatic summaries at the top of regular results, are a different surface with their own trigger patterns and citation rules; we cover those in our explainer on who gets cited in AI Overviews for HVAC searches. One term to pin before the data: throughout this post, a citation means a linked source inside an AI response — not the local-SEO sense of a directory listing.

01

What is AI Mode and who uses it?

AI Mode is Google's opt-in conversational search — a separate tab where a question gets a full AI-generated answer with supporting links, instead of the classic ranked results. Google began rolling it out to everyone in the US on May 20, 2025 at its I/O event, describing it as its "most powerful AI search, with more advanced reasoning and multimodality."

The adoption numbers come from Google itself, on the record. On the Q2 2025 earnings call, CEO Sundar Pichai put AI Mode at over 100 million monthly active users across the US and India — against 2 billion monthly users for AI Overviews. By the Q4 2025 earnings call, the usage pattern had sharpened: daily AI Mode queries per user in the US had doubled since launch, and Google shipped over 250 product launches across AI Mode and AI Overviews in that quarter alone.

Two behavioral details from those same remarks matter for a home-services business. First, queries in AI Mode are three times longer than traditional searches — people type "my AC is blowing warm air and the outdoor unit is making a clicking noise, what's wrong and who should I call" rather than "ac repair." Second, nearly one in six AI Mode queries are non-text — spoken or photographed. A homeowner pointing a camera at a frozen refrigerant line is a search your service pages never saw coming.

Google's own documentation for site owners describes where AI Mode fits: it is "particularly helpful for queries where further exploration, reasoning, or complex comparisons are needed." That is the research phase of hiring a contractor — comparing heat pump vs furnace replacement, weighing repair against replacement, shortlisting companies. The emergency "water heater burst" search still resolves fastest through classic results and the map pack. AI Mode is absorbing the deliberation end of home-services discovery first.

Here is the shape of the difference, using only what Google and the two largest vendor studies have published:

DimensionTraditional Google searchAI Mode
What the user seesRanked links, ads, map packOne AI answer + a sidebar of linked sources
Typical queryShort keywords3x longer, conversational, follow-ups common (Google, Q4 2025)
How sources are chosenRanking systems order pagesQuery fan-out retrieves pages across many sub-queries
Overlap with top-10 organic— (it is the top 10)~54% of domains, ~35% of URLs (Semrush, July 2025)
Local resultsMap pack on local-intent queriesLocal pack included on 13.49% of queries (Semrush)
Sources shown per answer10 organic resultsSidebar averages 7 unique domains (Semrush)
How you measure itSearch Console, by query and pageLumped into Search Console's overall "Web" type — no separate breakout (Google)
02

How AI Mode picks sources differently

AI Mode selects sources through query fan-out: it breaks a question into subtopics and, in Google's own words, issues "a multitude of queries simultaneously on your behalf", then assembles an answer from pages retrieved across all of those sub-queries. A traditional search runs one query and ranks pages against it. AI Mode effectively runs dozens, which is why its cited sources routinely include pages that never ranked for the original query.

Google's site-owner documentation confirms both AI Overviews and AI Mode may use fan-out, and adds the consequence in plain terms: the technique lets Google "display a wider and more diverse set of helpful links" than a classic search — and because the two features "may use different models and techniques," the links they show will vary.

The vendor data measures how far that variance goes. Semrush's July 2025 study of 5,000 keywords and 150,000+ citations found AI Mode's citations overlapped Google's top-10 organic results at roughly 54% at the domain level and 35% at the URL level — the loosest relationship of Google's own surfaces. For comparison:

PlatformDomain overlap with Google top 10URL overlap
Perplexityover 91%82%
Google AI Overviews~86%67%
Google AI Mode~54%~35%
ChatGPTlowest in the studylowest in the study

Source: Semrush AI Mode study, July 2025 — vendor research, not a Google-published metric.

Read the AI Mode row carefully: about half the domains it cites come from the top 10, but only about a third of the URLs do. AI Mode trusts many of the same websites and then reaches deeper into them — a specific troubleshooting article or FAQ page instead of the homepage that ranks. Semrush's format data shows the same split within AI Mode itself: 92% of responses carried a link sidebar averaging seven unique domains (versus three in AI Overviews), while the 7% of responses with additional links below the answer hewed close to classic rankings — 89% domain and 80% URL overlap.

The divergence is just as sharp against Google's other AI surface. Ahrefs analyzed 730,000 response pairs (September 2025 US data) where the same query produced both an AI Overview and an AI Mode response. The two agreed on substance — 86% semantic similarity — while citing the same URLs only 13.7% of the time. Only 16% of their words overlapped, and they opened with the same first sentence in just 2.51% of cases. Two Google systems, one question, near-identical conclusions, almost entirely different sources.

For a business, the entity findings in that study cut both ways. AI Mode responses run roughly 4x longer and mention 3.3 people or brand entities on average versus 1.3 in AI Overviews — more room for your company's name to appear. And when a brand earns a mention in an AI Overview, it also appears in the corresponding AI Mode response 61% of the time, but alongside additional competitors that didn't make the shorter answer's cut. More exposure, more company.

The honest summary: AI Mode neither mirrors the rankings nor ignores them. Domains that rank widely get retrieved more often — Semrush found a strong domain-level correlation between top-10 rankings and AI citations — but which page gets cited is decided by fan-out retrieval, not by position one.

03

What this does to the map pack

The map pack survives inside AI Mode — as one component of an AI answer rather than the centerpiece of the page. In Semrush's study, 13.49% of AI Mode queries included a local pack in the response. When a question has local commercial intent — "who repairs ductless mini splits near me" — AI Mode renders business listings inside the conversation, drawn from the same Business Profile data that powers the classic map pack.

That number deserves its qualifier: 13.49% comes from a 5,000-keyword sample spanning all query intents, not a local-services-only panel. Nobody has published an HVAC-specific AI Mode local-pack rate yet. What the number does establish is that Google carried its local infrastructure into AI Mode rather than replacing it — your Google Business Profile is a citation source for AI answers, not a legacy asset.

Google says as much in its own guidance. The SEO best practices list for AI features explicitly includes "checking that your Merchant Center and Business Profile information is up-to-date" — the only place a specific Google product gets named in that list. The reviews, categories, photos, and hours that win the classic map pack feed the same machine that populates AI Mode's local results. That work sits in our local SEO scope for HVAC companies precisely because it now pays on two surfaces at once.

The forward-looking piece is agentic. At I/O, Google announced agentic capabilities coming to AI Mode — starting with event tickets, restaurant reservations, and local appointments. Google named booking partners for the launch categories; home services was not among them. But the direction is explicit: AI Mode is being built to complete local transactions, not just describe them. A contractor whose booking flow only works for a human filling out a form is worth stress-testing against that trajectory.

What the data does not support: claims that AI Mode has killed the map pack. Emergency and "near me" searches still run through classic results, where the pack still renders. The measured change is additive — a second, conversational surface where local visibility gets decided by different mechanics.

04

Early data on AI Mode and local businesses

The earliest reliable finding is speed with volatility. Semrush published 81 new pages in September 2025 and tracked when AI platforms cited them. Google AI Mode cited 29 of the 81 pages (36%) within 24 hours of publishing — more than three times faster than ChatGPT search. Then the churn started: the cited count climbed to 48 pages, fluctuated almost daily, and fell to 21 pages (26%) by day 30. ChatGPT moved opposite — 8 pages on day one, growing steadily to 34 (42%) by day 30, with citations that tended to stick. (How the non-Google engines find and weigh local businesses differs enough that we compare them separately in Perplexity vs ChatGPT for local business visibility.)

Two caveats keep that finding honest. The test site carried an Authority Score of 84 — an established domain with years of trust; a contractor site should expect slower pickup, not 24-hour citations. And the volatility matches what Ahrefs found on Google's other AI surface, where 45% of AI Overview citations changed between generations of the same answer. An AI Mode citation is a rolling audition, not a ranking you hold.

Three more early patterns bear directly on home-services visibility:

  • Commercial questions get the long answers. Across the platforms Semrush tested, commercial and transactional queries triggered responses roughly twice as long as informational ones. "Best HVAC company for a heat pump install" produces a fuller, more source-hungry answer than "what is a heat pump" — which means more citation slots on exactly the queries that precede a purchase.
  • Third-party surfaces carry unusual weight. Reddit, YouTube, and Facebook appeared in over 68% of AI Mode results that included additional links in Semrush's data. What homeowners say about contractors on community platforms is raw material for AI Mode answers — a reputation layer no service page substitutes for.
  • Longer answers surface more businesses. Ahrefs measured 34.66% of AI Mode responses containing no brand entities at all, versus 59.41% for AI Overviews. AI Mode names companies almost twice as often — being nameable, consistently, across the web is what converts that into visibility.

What does not yet exist anywhere, from any vendor: published data on how often AI Mode cites HVAC contractors specifically, or which sources it pulls for HVAC service queries. That gap is why we run our own recurring sampling of AI answers for HVAC queries — the HVAC AI visibility report we are building from that panel will publish the numbers rather than assert them. Until trade-specific data exists, the studies above are the floor you can plan on, and anyone quoting an "HVAC AI Mode citation rate" should be asked where it came from.

05

Preparing an HVAC site for AI Mode

Preparation starts from one fact Google states outright: there are no additional technical requirements for AI Mode. A page must be indexed and eligible to show in Search with a snippet — the same gate as classic results. Everything beyond that gate is probability-raising, and here is the order we'd run it for an HVAC contractor:

  1. Confirm the eligibility gate. Indexed, snippet-eligible, no stray noindex or nosnippet on service pages. If a page can't appear in classic search, no AI surface can cite it — full stop.
  2. Clear the crawl path, including your CDN. Google's best-practices list leads with allowing crawling "in robots.txt, and by any CDN or hosting infrastructure." Firewall defaults block bots silently; our five-minute crawler access test covers the same failure layer for every AI crawler family, Googlebot included.
  3. Build pages that answer the fan-out's sub-questions. AI Mode decomposes "should I repair or replace a 12-year-old AC" into cost, lifespan, refrigerant, and sizing sub-queries. A page that answers one of those directly — heading, straight answer, evidence — is retrievable for it; a thin overview page answers none of them. This is where the three-times-longer conversational query becomes an advantage for specific content.
  4. Keep the Business Profile current. Google's own AI-features guidance names it; the 13.49% local-pack rate shows why. Categories, services, hours, photos, reviews — the profile is now source data for AI answers.
  5. Earn a presence where AI Mode already looks. Reddit threads, YouTube, community groups — the 68%+ figure above. Real participation and real reviews, not planted mentions; businesses earn citations from the footprint others write about them.
  6. Measure it honestly. Search Console reports AI Mode clicks inside the overall "Web" type with no breakout, so first-party attribution is genuinely limited. Recurring sampled prompts, tracked over time, are the working method — directional evidence, reported as such. That measurement discipline is the backbone of how we build AI-answer visibility for HVAC companies: access audit, answer-shaped content, profile and reputation work, and a citation-sampling loop — sold as a program with measurement, never as a promised placement.

And the line we put in writing because the volatility data demands it: nobody controls AI Mode's output. A vendor guaranteeing your company a spot in AI answers is guaranteeing a system that re-decides its sources between generations of the same answer. The evidence supports raising your odds; it does not support promises.

FAQ

Frequently asked questions

Do I need separate SEO for AI Mode?

The foundation is identical — Google states the same SEO best practices and technical requirements apply, with nothing extra. But the citation data argues for tracking it separately: AI Mode's sources overlap AI Overviews' only 13.7% of the time and top-10 organic URLs about 35%, so visibility on one surface doesn't confirm the others.

Does ranking #1 on Google get me cited in AI Mode?

It helps and doesn't guarantee. Semrush found strong domain-level correlation — sites with many top-10 rankings earn more AI citations — but only ~35% of AI Mode's cited URLs come from the top 10 for that query. AI Mode often cites a deeper page from a ranking domain rather than the page that ranks.

Does AI Mode show a map pack?

Yes, on a minority of queries — 13.49% in Semrush's 5,000-keyword study included a local pack inside the AI Mode response. The listings draw on Google Business Profile data, which is why Google's AI-features guidance explicitly recommends keeping your profile up-to-date.

Does llms.txt help my site appear in AI Mode?

No. AI Mode eligibility runs entirely through Googlebot's index, and Google ignores llms.txt — it is not a Google ranking factor. The file's actual (modest) role with other AI engines is a separate question, covered in our llms.txt guide for contractor websites.

Can an agency guarantee my HVAC company appears in AI Mode?

No — and the volatility data makes the guarantee itself a red flag. Semrush watched AI Mode's cited pages swing from 36% to a peak near 59% and down to 26% within 30 days on a high-authority domain. Honest vendors sell access audits, answer-shaped content, and recurring measurement, not placements.

Want this working for your HVAC company?

Transparent pricing, month-to-month, and a plan built from your market's data — not a template.