A homeowner in Gainesville wants to know what a new HVAC system should cost before she calls anyone. Two years ago she typed that into Google and clicked through three or four articles. Today she gets a synthesized answer at the top of the page, written by AI, assembled from sources she never visits.
If your business was one of those sources, you may still get the call. If it wasn't, you were never in the conversation.
This is the shift every service business is being warned about, usually badly. The panic version says AI is going to erase your website. The dismissive version says nothing has really changed. Both are wrong, and the actual picture is far more useful once you separate two very different kinds of searches.
The distinction that changes everything
Not all searches are being absorbed by AI at the same rate. This is the single most important thing to understand, and almost nobody leads with it.
Informational searches are being heavily reshaped. "How much does AC replacement cost," "how long should a roof last in Georgia," "signs my water heater is failing" — these are being answered directly. Industry tracking put AI Overviews on roughly 48% of all Google searches as of March 2026, and where they appear, one widely cited study found organic click-through rates drop by about 38%. Those are your top-of-funnel blog posts and guide pages. They are absolutely under pressure.
Local-intent searches are largely intact. "AC repair near me," "roofer Cumming GA," "emergency plumber Dahlonega" — these still return a map pack with businesses, phone numbers, star ratings and directions. One analysis found AI Overviews appear in only about 7% of local searches. Google knows that someone searching "emergency plumber" wants a phone number, not an essay.
Sit with the gap between 48% and 7% for a second, because it's the whole strategy.
Your bottom-of-funnel, ready-to-buy traffic is far safer than the headlines suggest. Your research-phase traffic is genuinely being disrupted. Which means the correct response is not to abandon content — it's to change what your content is for. It used to earn a click. Now it also has to earn a citation.
What AI is actually reading when it picks a business to name
When someone asks an AI assistant "who should I call for a leaking water heater in Dawsonville," the system isn't running a popularity contest. It's assembling an answer from entity signals — the accumulated, consistent, machine-readable evidence of what your business is, where it operates, and how well it performs.
Four things carry that weight.
Consistent identity data. Your name, address and phone number have to match across your website, your Google Business Profile, and the directories that feed the wider web. AI systems cross-reference. Contradictory data doesn't just rank lower, it lowers confidence that you're a real, verifiable entity — and low-confidence entities don't get named in answers.
Explicit, structured service descriptions. "We do it all, no job too small" is invisible to a language model trying to match a specific problem. "We repair and replace tankless water heaters in Dawson, Lumpkin, Hall and Forsyth counties, typically same-day" is legible. Say the actual service, in plain words, on a page dedicated to it.
Review text, not review scores. This deserves its own section. See below.
Schema markup. LocalBusiness, Service, and AggregateRating schema hand machines a labeled version of facts they would otherwise have to infer from your layout. It's the difference between leaving a business card and leaving a filled-out form.
Your reviews became training data
Here is the change most business owners have not registered yet.
A five-star review with no text has almost no value to an AI system. It's a number with no content. But this review:
"Replaced our tankless water heater the same afternoon we called. The tech walked through both options, explained why the cheaper one wouldn't handle our square footage, and the final price matched the quote exactly."
That paragraph tells a language model the service performed, the response time, the consultative approach, and the pricing integrity — in natural language a real customer used. When someone later asks an AI "who does same-day water heater replacement and gives honest quotes," that review is the evidence that gets you named.
Your reviews stopped being social proof for humans and became structured evidence for machines. Same reviews, second job.
Which means your review request needs to change. Stop asking for stars. Ask a question that produces a sentence: "Would you mind mentioning what we fixed and how the process went?" One line of difference in your follow-up text, and the reviews you collect start doing twice the work.
Rebuild your content around problem states, not services
The content that survives this shift is the content that answers the question a customer actually types before they know who to call.
Most service business websites are organized around what the company sells. "Our Services. HVAC Installation. Duct Cleaning. Maintenance Plans." That structure is a filing cabinet. It matches almost nothing anyone searches.
Customers search from inside a problem: my upstairs won't cool down, my bill jumped $180 this month, there's water pooling around the furnace, my roof is only nine years old and there are granules in the gutter. Build pages for those. Answer the question completely and honestly — including the cases where the answer is "this one's usually a cheap fix, here's how to do it yourself."
That last part feels wrong to owners and it's the highest-ROI thing on this list. Content that is genuinely useful gets cited by AI systems and trusted by readers. Content that withholds the answer to force a phone call gets skipped by both.
Structure it so machines can lift it cleanly: a direct answer in the first two or three sentences, then the depth. Clear question-style headings. Short paragraphs. Real numbers and ranges where you can give them.
What this doesn't change
Two things stay exactly as they were, and it's worth saying plainly because the AI conversation has gotten loud enough to distract people from fundamentals.
The map pack still decides local. Yesterday's post covered it — categories, proximity, review recency, profile completeness. That system is running the same way it was, and it still governs the searches with the highest purchase intent. AI didn't replace it. If you have to pick one thing to fix this month, it's still that.
Answering the phone still closes the job. No amount of AI visibility survives an unanswered call. Tomorrow's post has the numbers.
The move
Get named in the answer instead of arguing with the algorithm:
- Clean your entity data — NAP consistency across your site, GBP, and top directories.
- Add LocalBusiness, Service and AggregateRating schema so machines read facts instead of inferring them.
- Rewrite your review request to ask for a described experience rather than a rating.
- Build problem-state pages for the eight or ten questions customers actually ask before they call.
- Lead every page with the answer, then explain — the structure AI systems extract from and readers prefer anyway.
- Keep the map pack strong, because that's where local intent still lands.
None of this is exotic. It's the same discipline good local SEO always required, aimed at a reader that happens to be a machine.
Curious what AI actually says about your business today? AAMG runs a visibility check across AI search, traditional organic, and the local pack, then shows you exactly where the gaps are. Schedule a Strategy Call.
AAMG builds the marketing, technology, and automation systems businesses need to grow. Build smarter. Market better. Scale further.
