How AI search is replacing Google for local services — by the numbers
Forty-five percent of homeowners now ask AI apps to recommend local services. A year ago that number was six percent. The shift, the data, and what it means for an independent HVAC contractor in 2026.
Three years ago a homeowner with a dead AC opened Google, typed "AC repair near me," and got ten map results plus ten links. Your business, if it ranked, was one of them. Today that same homeowner opens ChatGPT or Google's new AI assistant and types the same thing. Instead of ten options the AI picks one. Sometimes three. And the algorithm that picks isn't anything like Google's old PageRank — it's based on what the model has learned the open internet "knows" about your business across reviews, news mentions, directories, structured data, and the conversational tone of pages that mention you.
Most independent HVAC contractors weren't built for that kind of recognition. They were built for the local map pack and a handful of paid clicks. The recognition AI assistants reward — entity strength, NAP consistency, citation distribution across directories, schema markup — wasn't on anyone's checklist five years ago. The contractors who figured it out (a small minority) are pulling away. The ones who didn't are losing booked jobs they don't even know they were considered for.
Where did the 6%-to-45% number come from?
The 45% figure is from MarketingCode's 2026 US Local Services Search Behavior Study, which surveyed 4,200 US homeowners about how they researched and chose service providers in the prior 90 days. The 6% baseline is the same study's 2025 reading. MarketingCode publishes the study annually and it has become one of the most cited data points in local-search trade press for a reason: the methodology controls for age, income, region, and prior tech adoption. The behavior shift is real and it is national, not concentrated in coastal metros.
The same survey breaks the 45% down by intent. About 28% of those AI-first searches are general ("who do I call for HVAC near me"), 11% are comparison-driven ("is Roto-Rooter or a local company better"), and 6% are job-specific ("AC company that does ductless mini split installs in {city}"). Comparison and job-specific queries are the ones independents win — if they show up at all.
Why did the shift happen so fast?
Three things converged in 2025 and 2026. First, OpenAI added free web-search to ChatGPT for all users, removing the friction of subscribing or installing a separate browser. Second, Google switched on AI Overviews for most local-services queries — the box at the top of search results that summarizes an answer rather than just listing links. Third, Anthropic released Claude for Small Business and ran a 10-city US tour explicitly teaching SMB owners and consumers how to use AI for everyday research. The combined effect was a measurable monthly increase in the share of "who do I call" intent that resolves inside an AI app instead of a search engine.
Anthropic published the small-business tour cities and adoption metrics on their company blog (anthropic.com) in May 2026. The pattern there is consistent with the MarketingCode survey: adoption is fastest among 35-55 year old homeowners — the exact band that buys HVAC service.
Who wins inside the AI answer?
When we ran our own audit on a sample of 30 standardized homeowner-style queries against four AI engines (ChatGPT, Claude, Perplexity, Google AI Overviews), the same three brand names appeared again and again:
Source: TNova Labs internal audit, 30 queries × 4 AI engines
That last bar is misleading on its own. "All other independents combined" sounds healthy, but the 80.8% is split across literally hundreds of named businesses, with a long tail where most contractors get 0% — they aren't named at all. The chains' 19% combined share is concentrated. The independents' 81% is fragmented to the point of noise.
This is the math that matters: if your business has 1,200 reviews and 30 years in the trade and you get 0% citation share, your reputation is invisible to the customer who asked the AI. The customer is told to call Roto-Rooter, who has 0 reviews of work in your specific city, because Roto-Rooter has the entity strength — the structured presence across directories, news, listings, and consistent NAP — that the AI weights heavily.
What does "entity strength" actually mean?
Entity strength is the umbrella term for how recognizable a business is to a knowledge-graph-driven AI system. The components are roughly:
1. NAP consistency — your name, address, and phone number matching byte-for-byte across Google Business Profile, Yelp, BBB, Yellow Pages, Angi, HomeAdvisor, and at least eight other directories.
2. Citation distribution — how many distinct domains mention your business (not just rank you), and the authority of those domains.
3. Structured data — schema.org markup on your website (HVACBusiness, LocalBusiness, FAQPage) that machines can parse, not just humans.
4. Review velocity — not raw count, but how recently and how steadily new reviews arrive, ideally with operator responses.
5. Mentions in trade press, local news, and industry directories — the kind of editorial signal that AI engines treat as third-party validation.
Roto-Rooter wins on items 2, 4, and 5 because they spend marketing dollars on press releases, paid placements in directories, and franchise-level review-acquisition systems. They win on item 1 because they have a centralized listings team. They win on item 3 because their corporate web team understands schema.
An independent contractor can win on items 1, 3, and 4 with focus and a small budget. Items 2 and 5 take longer. The point of starting now is that the curve compounds — every month of NAP cleanup and consistent review velocity accrues, and AI engines reward that consistency more than they reward last week's spike.
So what changes in your day-to-day?
Probably less than you think. The contractor doesn't need to learn schema markup or run their own AI audit. The contractor needs to: (a) have a current website that a vendor can clean up, (b) verify their Google Business Profile claim and complete every field, (c) ask satisfied customers for reviews steadily, and (d) be willing to fix or hire someone to fix the technical pieces underneath. The work that moves the AI score the most is invisible to the homeowner — it happens in directories, in JSON-LD blocks, in citation databases. But it's what determines whether your business is the answer the next time a homeowner asks ChatGPT for HVAC help in your city.
What to expect in the next 12 months
If the MarketingCode trend continues at its current rate, AI-first local-services search will pass 60% of all homeowner queries by the end of 2027. The chains will continue to widen their lead inside the AI answer unless independents start investing in the underlying entity strength now. The window where the independent contractor can still win on specificity ("best HVAC for older homes in {city}", "family-owned HVAC near me") is open today and will narrow as more contractors wake up to the shift.
We built TNova Labs to do exactly that work for HVAC contractors who don't want to learn what schema is. The audit is free, the report is yours to keep, and we bill on outcomes — when your phone actually rings, not for activity reports. The link is at the bottom of this post.
Is the 45% number really US-wide or just a coastal-metro phenomenon?
MarketingCode's 2026 sample (n=4,200) is geographically weighted to the US Census distribution, so it's national. The shift is happening fastest in metros above 1M population but the trend line is the same in smaller markets, just trailing by 6-8 months.
Does this mean Google Ads doesn't matter anymore?
No. Google Local Services Ads and Google Search Ads still produce booked jobs and will for the foreseeable future. The point is that AI-first search is now a parallel acquisition channel that is invisible to a marketing dashboard built only around paid clicks. Both channels need attention.
Why do the chains dominate the AI answers?
Because AI engines reward entity strength — structured presence across directories, news mentions, consistent NAP, schema markup, and review velocity. The chains have centralized teams that maintain those signals at scale. Independents can match or beat them on specificity (older-home expertise, ductless installs, financing options) but only after the underlying entity work is done.
How long does it take to move the citation share number?
In our audits, the first measurable lift typically arrives at week 6-8 after listings cleanup and schema work are deployed. Compounding lift continues for 9-12 months as the changes propagate through directory caches and AI training refresh cycles.
Can I do this myself without hiring an agency?
Yes, on the easy items: claim your Google Business Profile, complete every field, ask for reviews steadily, and audit your NAP across the top 6 directories. The harder items — schema markup, llms.txt, citation distribution across 50+ directories, and tracking changes in citation share monthly — are technical work that a contractor without a marketing team will not realistically maintain. That's where an agency earns its keep.
What is an AI citation score?
It's a measurement of how often a business appears in AI-generated answers when homeowners ask about local services. It's expressed as a 0-100 number computed from your share across multiple AI engines for a standardized set of homeowner-style queries. We publish our own scoring methodology at /methodology and the audit itself is at /visibility-audit.
See where you show up when customers search.
Free 24-hour Visibility Audit. Yours to keep — whether you hire us or not.