TL;DR
Google AI Overviews now appear in roughly 48 to 55 percent of searches. ChatGPT, Perplexity, and Claude are pulling local leads before your website ever loads. Service businesses that do not structure their content for machines will lose calls to competitors who do. This is not a ranking problem. It is a citation problem. If an AI cannot read your page, it cannot cite your page. If it cannot cite you, you do not exist in the answer. Below are seven steps: FAQ schema, direct-answer content, Google Business Profile optimization, location pages with structured data, third-party citation building on Yelp and Angi and BBB, People Also Ask content clusters, and the llms.txt file. Do all seven. Systems beat slogans.
The Bridge Just Moved
I ran nuclear reactors on a submarine before I ran a marketing company. On a boat, you do not get to argue with the instrument panel. The gauge reads what it reads. You adjust your systems to match reality, not the other way around.
Search just handed every service business a new instrument panel, and most owners have not looked at it yet.
Google AI Overviews are showing up in 48 to 55 percent of searches as of early-to-mid 2026, depending on which tracker you trust. Some trackers put personalized U.S. results north of 60 percent. The exact number moves month to month as Google recalibrates. The direction does not move. It only goes up. ChatGPT is processing 2.5 billion queries a day, and AI referral traffic converts at roughly five times the rate of traditional organic clicks, because the person asking already trusts the answer they got.
Here is the part that should get your attention if you run a plumbing company, a law firm, an HVAC outfit, or a dental practice. Local searches are still mostly untouched by AI Overviews inside Google itself, only around 7 percent trigger one, because Google still trusts its Map Pack for "near me" intent. But that stat is misleading comfort. The buyer is not always searching inside Google anymore. They are asking ChatGPT "who's the best plumber in Austin" and getting a named answer, not a list of ten blue links. Perplexity does the same. So does Gemini when it operates outside classic Google Search.
Answer Engine Optimization, AEO, is the discipline of making sure your business is the one named. Not ranked. Named. That is a different game with different rules, and most service business owners are still playing the old one.
The Owner-Operator Frame
On a sub, every system exists to answer one question: does the ship stay safe and functional under pressure. Not "does it look good on an inspection." Functional under pressure.
Apply the Owner-Operator Frame to your website the same way. An Owner-Operator does not ask "is my content good." An Owner-Operator asks "can the machine reading this content extract an answer, verify who I am, and hand my business to a customer with confidence." That is a different bar. A blog post can read beautifully to a human and stay invisible to an AI system, because it never answers the question directly, never carries schema, and never gets corroborated anywhere else on the web.
Owner-Operators build for the machine first, because the machine now stands between you and half your incoming leads. The human still reads what the machine hands them. The machine decides whether you get handed over at all.
When I launched demg.ai, one of the first things we built was structured data on every page. Not because it was trendy. Because if an AI cannot read your content, it cannot cite your content. And if it cannot cite you, you do not exist in the answer.
That is the whole doctrine in one sentence. Everything below is the checklist version of it.
Step 1: Add FAQ Schema to Every Service Page
FAQPage schema is the single cheapest, highest-impact move on this list. It tells an answer engine, in its own language: this is a question, this is the accepted answer, this is how the two connect. Research from HubSpot and CXL both confirm FAQPage schema is a core AEO signal alongside LocalBusiness, Article, and Speakable markup.
Do not fake this with generic filler questions. AEO researchers at WebFX found that adding FAQ schema to weak, generic answers does nothing. Schema clarifies content. It does not rescue thin content. Write real questions your customers actually ask ("How much does a water heater replacement cost in Dallas?") and answer them directly in the first sentence.
Validate every implementation with Google's Rich Results Test. Broken schema sends conflicting signals to AI systems, and conflicting signals get you skipped, not just ignored.
Step 2: Write Direct-Answer Content
Question as the H2. Answer in the first sentence, 40 to 60 words, no throat-clearing. This is the exact format cited across CXL's and Surmado's AEO research: place a complete, direct answer immediately after the question heading, then support it with bullets or a short paragraph underneath.
Compare these two openings for "How much does AC installation cost in Houston?"
Weak: "There are many factors that go into the cost of AC installation, and it's important to understand your options before making a decision."
Strong: "AC installation in Houston typically costs $4,500 to $9,000 depending on unit size and ductwork condition. Most homeowners pay $6,200 on average for a 3-ton system."
The second version is extractable. An AI can lift that sentence, cite you, and hand it to the user. The first version forces the AI to guess or move to a competitor's page that actually answers the question. Long-tail, multi-word queries are roughly seven times more likely to trigger an AI Overview than short ones, which means the specific, detailed answer does more work than the generic one, not less.
Step 3: Optimize Google Business Profile for AI Readability
Your Google Business Profile is not a listing anymore. It is a data feed that AI systems crawl to verify you exist. HubSpot's AEO research is blunt about this: complete hours including holidays, high-quality photos, every relevant category selected, and a keyword-rich business description that mirrors the language customers actually search.
Write the description the way a customer would ask a question, not the way a brochure would describe you. "24-hour emergency plumbing serving Bellevue, Redmond, and Kirkland with licensed technicians and upfront pricing" gives an AI system specific service-area and differentiation data to extract. "Full-service plumbing company" gives it nothing.
Step 4: Build Location-Specific Landing Pages With Structured Data
One page that says "serving the greater metro area" is worth less than five pages, each with the city name in the URL, the H1, the meta description, and the schema's areaServed field. This matters because AI engines need geographic clarity to match a user's location to your service area with confidence.
Each page needs LocalBusiness or Service schema with the city named explicitly. Split broad categories into specific services too. "Plumbing services" is vague. "Drain cleaning in Redmond," "water heater repair in Bellevue," and "leak detection in Kirkland" each give the AI a clean, citable unit of meaning.
Step 5: Get Cited in Industry Directories
This is the step most owners skip, and it is the step the data says matters most. A study analyzing 28 million AI responses to local business queries found Yelp receives 3.4 times more AI citations than the next closest local discovery platform, and more citations than the other five major platforms combined, over 512,000 citations in a single quarter. More than half of those citations came from "near me" queries specifically. A separate analysis of 267,000 AI citation mentions across active local campaigns confirmed the pattern: Yelp and Google dominate, with Angi and BBB providing category-specific authority that AI systems trust for vertical accuracy.
The mechanism matters more than the platform. AI engines weight a distributed review presence higher than a concentrated one. Fifty reviews spread across Google, Yelp, and an industry-specific directory like Angi or Houzz reads as "established across the category." Three hundred reviews on Google alone and nothing else reads as either review-pumping or a single-channel operation. Name, address, and phone data has to match exactly across every platform too. A single phone number discrepancy between your website and your Yelp listing is enough to suppress your citations, because inconsistent data creates uncertainty, and uncertain AI tools name someone else.
Step 6: Create People Also Ask Content Clusters
Build a dedicated "Questions We're Asked" hub, five to ten real customer questions per service, each answered in 40 to 60 words immediately under the heading, then supported with bullets or a short paragraph. This is not a single FAQ page bolted onto your homepage. It is a cluster: pull the actual "People Also Ask" queries Google surfaces for your core services, and build a page or section for each one.
Add one expert quote or specific data point per cluster where you can. Princeton's GEO research, cited across multiple AEO guides this year, found that inline citations, specific statistics, and attributed expert commentary meaningfully increase the odds a page gets pulled into a generated answer. Generic paragraph content optimized for keyword density does almost nothing here. Question-and-answer structured content is what gets extracted and cited.
Step 7: Submit to llm.txt and ai.txt Standards
I will give it to you straight, because Dan Kennedy trained me not to sell you a fantasy. llm.txt adoption grew nearly nine-fold in the past year, but independent server-log research found 97 percent of llms.txt files received zero AI requests as of May 2026. Not one of the major consumer AI platforms, ChatGPT, Claude, or Gemini, currently serves a validated llms.txt on their own consumer-facing domains. Google has stated publicly it does not affect Search rankings either way.
So why is this step seven and not step zero. Two reasons. First, three real crawlers, OpenAI's search bot, PerplexityBot, and Anthropic's ClaudeBot, do fetch llms.txt regularly, and sites that ship it report a measurable citation lift from those three engines specifically within about two weeks. Second, the file costs fifteen minutes to build: a curated list of your ten to thirty most important URLs with one-sentence summaries at your domain root. The downside of skipping it is close to zero. The upside, if adoption accelerates the way robots.txt did after 1994, is a clean head start. Ship it. Do not restructure your whole content strategy around it.
Doctrine Connection: Systems Beat Slogans
None of these seven steps work in isolation, and that is the point. Schema without direct-answer content is decoration. Direct-answer content without directory citations is an island. Directory citations without consistent NAP data across every platform actively confuse the machine you are trying to convince.
A slogan is "we're AI-optimized now." A system is FAQ schema on every service page, validated against Google's Rich Results Test, feeding direct-answer content that mirrors real customer questions, anchored by a Google Business Profile that matches your Yelp listing character for character, corroborated by four to seven citations on trade-relevant directories, organized into People Also Ask clusters, with a fifteen-minute llms.txt file at the root for good measure.
Research tracking home services AI visibility found the businesses that win are not the ones with the single best signal. They are the ones balanced across all five categories of trust, strong in at least two. A business with strong reviews and weak press loses to a competitor with moderate reviews and four press mentions. The recommendation engine treats you as a portfolio of signals, not a single profile. Build the portfolio. Systems beat slogans, every time, because a slogan cannot be crawled and a system can.
FAQ
What is Answer Engine Optimization and how is it different from SEO? SEO optimizes for ranking in a list of results. AEO optimizes for being the direct answer itself, the thing an AI names when a user asks a question instead of a list of links they have to click through. AEO depends on structured data, direct-answer formatting, and consistent citations across trusted third-party sources rather than keyword density and backlinks alone.
Which schema type matters most for a local service business? LocalBusiness schema is the foundational signal, paired with Service schema on each individual service page and Review schema wherever testimonials appear. FAQPage schema is the next priority. Businesses with a full schema stack get cited at roughly three times the rate of competitors running no schema at all, according to trust-signal research across service categories.
Do I need to be on Yelp if my customers mostly come from Google? Yes. AI platforms build local recommendations from a source graph that includes Yelp, Reddit, Google, Angi, and BBB, and Yelp specifically receives more AI citations than every other local discovery platform combined. Even customers who found you through Google are increasingly asking ChatGPT or Perplexity for a second opinion, and those engines lean heavily on Yelp data to answer.
How long does it take to see results from AEO work? Structured data and FAQ content can influence AI citations within two to four weeks. Building a real directory citation footprint across four to seven authoritative sources typically takes four to eight weeks of consistent work. Most operators see full AEO visibility consolidate somewhere between sixty and ninety days, assuming the underlying content and business data were solid to begin with.
Is llms.txt worth doing right now? It is worth fifteen minutes, not a redesign. Adoption is climbing fast but actual usage by the major consumer AI platforms is still close to zero, and Google has said it does not affect Search rankings. Ship a minimal version because the downside is nothing and three real crawlers do use it, but do not treat it as a priority ahead of schema, content, or citations.
*Jeff Barnes, MBA has no personal position in any company, fund, or platform named in this article. demg.ai has no current commercial relationship with any party mentioned. demg.ai provides marketing education and systems consulting, not investment advice. Past performance does not guarantee future results.*