A meaningful share of the people who used to type "hvac repair near me" into a search box now ask an assistant instead, and get back a short synthesized answer naming two or three businesses.
If you are not one of the named businesses, you are not in a lower position. You are absent. There is no page two to be on.
The good news is that most of the work that gets you cited is work you should already be doing. The part that is genuinely new is smaller than the volume of commentary suggests, and it is worth separating the two so you do not get sold an expensive solution to a problem that ordinary competence solves.
What is actually different
Traditional search returns a list and the searcher chooses. An assistant reads sources, synthesizes, and returns an answer, usually naming a small number of businesses, often with citations.
Three consequences follow.
The result set is much smaller. Ranking eighth on a results page still gets you traffic. Being the eighth-best source for a synthesized answer gets you nothing.
Consistency matters more than it did. A model assembling an answer from several sources resolves conflicts between them. If your hours, service area, phone number, or service list differ between your site, your Google Business Profile, and directory listings, you are a source that contradicts itself. That is a bad position when something is deciding which claim to repeat.
The question is longer. People ask assistants full questions — "who does emergency AC repair in Lakeland on a Sunday and how fast can they get here" — rather than typing two keywords. Content that answers a specific question directly has an advantage over content optimized for a short head term.
What has not changed
Most of it.
Assistants read the web. The sources they read most are the same authoritative, well-structured, crawlable pages that rank conventionally. There is no separate index you can buy into and no submission process that matters. Anyone selling you guaranteed placement in AI answers is selling something they cannot deliver.
Your Google Business Profile still matters enormously for local queries. Reviews still matter. A fast, crawlable site still matters. Being genuinely the best answer to the question still matters most.
If your local SEO fundamentals are weak, fix those first. There is no AI-specific tactic that compensates for a business that is hard to find, inconsistent about its own details, and thin on reviews.
The work that actually moves this
Make your operational facts unambiguous and identical everywhere. Service area, hours, emergency availability, services offered, pricing structure, phone number, business name. Identical on your site, your Google Business Profile, and every directory you appear in. This is dull and it is the highest-leverage item on the list, because it is where most businesses are quietly contradicting themselves.
Answer real questions in plain language. Take the questions you actually get on the phone and answer them on the site, directly, in the words customers use. Not "Comprehensive HVAC Solutions" — "Do you charge a diagnostic fee for emergency calls?" A page that answers a specific question completely is far more extractable than a page of positioning copy.
Put the answer near the top. Lead with the direct answer, then explain. Content that buries the answer under four paragraphs of preamble is harder to extract, and it is worse for human readers too.
Use structured data properly. LocalBusiness, Service, and FAQPage schema make your facts machine-readable rather than something to be inferred from prose. Mark up what is genuinely on the page — schema describing content that is not there is a liability, not an advantage.
Be citable from elsewhere. Assistants weigh what independent sources say about you. Local press, supplier and manufacturer directories, chamber and trade association listings, genuine partner pages. This is conventional local link building, and it is still the hardest and most durable part.
Keep pages fast and crawlable. Content that requires JavaScript execution to appear is less reliably read. Server-rendered HTML remains the safe default.
What to skip
Stuffing pages with question-shaped headings you do not answer. Sixty FAQ entries with one-sentence non-answers is thin content with a question mark on it.
Buying "AI visibility" packages that will not explain their method. Ask exactly what they will change on your site. If the answer is vague, the deliverable is vague.
Generating hundreds of near-identical city pages. This was already a weak strategy. Something that synthesizes across sources is, if anything, better at spotting that twelve of your pages are one page with the city name swapped.
Chasing every new metric dashboard. Measurement here is genuinely immature. Be skeptical of any tool claiming precise share-of-voice in AI answers.
How to know if it is working
Measurement is the weakest part of this discipline right now. Practical proxies:
Ask the assistants directly. Run the ten questions a customer would actually ask about your service and city, monthly, and record whether you are named. It is manual and it is the most honest signal available.
Watch referral traffic from assistant domains in your analytics. Volume is typically small relative to search, but the trend line matters and the intent is usually high.
Ask new customers how they found you, and give whoever answers the phone somewhere to record it. Self-reported attribution is imperfect and it will still tell you about a shift months before your analytics does.
The realistic framing
For most local service businesses this is not an emergency and it is not a new discipline. It is a reason to do the fundamentals properly: be genuinely findable, be consistent about your own facts, answer real questions in plain language, and earn mentions from sources that are not you.
Businesses that did that work are already getting cited. Businesses that skipped it are discovering that the shortcuts which produced mediocre rankings produce no citations at all.