AI Visibility for Service Businesses: What Actually Determines Who Gets Recommended

A year ago, 6% of consumers used AI tools to find local services. Today it is 45%. Google AI Overviews appear on roughly half of all U.S. searches and have cut click-through rates by up to 61%. ChatGPT has 900 million weekly active users. Among millennials and Gen Z, more than half no longer default to Google when they need to hire someone.

This is not a trend to watch. It is an infrastructure shift that has already happened. And for the vast majority of service businesses, the infrastructure they have built over the past fifteen years does not work for it.

This post is the full picture: how AI recommendations actually work, what data they rely on, why most businesses are invisible, and what to do about it. If you operate in a specific vertical, the deep dives at the bottom break this down by industry.

The shift is structural, not incremental

The standard reaction to declining organic traffic is to invest more in SEO. That diagnosis is wrong most of the time now.

When Seer Interactive tracked 3,119 queries across 42 organizations, organic CTR on queries with an AI Overview dropped from 1.76% to 0.61%. That is a 61% decline. You can hold Position 1 on Google and still lose more than half your clicks because Google answered the question above your link.

Meanwhile, the searches that never reach Google at all are growing faster. A homeowner asks ChatGPT who to call for a leaking water heater. She gets a name. She calls it. Your analytics show nothing. No referral source. No click. No impression.

Two channels are compressing simultaneously. Google is absorbing clicks at the top. AI chat is absorbing searches at the bottom. The businesses caught in between are the ones without a data strategy for either.

How AI recommendations actually work

AI does not rank websites. There is no PageRank equivalent. No keyword matching. No ad auction.

When someone asks ChatGPT, Perplexity, Gemini, or Claude to recommend a service provider, the model assembles an answer from three sources:

Training data. A static snapshot of the web, typically months old. If your business information hasn't changed recently, this is what the model knows about you. If your information has changed, the model doesn't know.

Retrieved context. At inference time, some platforms retrieve current information from indexed sources. JSON-LD markup on web pages, structured databases, indexed feeds. This is the real-time layer. Ahrefs found that 76.4% of the most-cited pages had been updated within the last 30 days. Freshness is not a nice-to-have. It is the single strongest predictor of citation.

Source reliability signals. AI systems face the same problem Google faced in 1998: self-asserted claims are unreliable. A business saying "we complete 3,000 jobs per year" on its homepage is an assertion. The same metric verified by a third party from authenticated accounting data is evidence. The model weighs verified, third-party-corroborated data more heavily than self-reported claims.

The result: the model is not choosing the best business. It is choosing the most evaluable one. The business with the most structured, verified, machine-readable data is the one the model can confidently name. Everything else gets skipped or described in vague generalities.

The data hierarchy

Not all data carries equal weight in AI evaluation. Based on how AI systems retrieve, compare, and cite information about service businesses, there is a clear hierarchy.

Tier 1: Verified operational metrics

Job volume. Repeat customer rate. Service mix distribution. Average customer tenure. Revenue consistency. Service area by actual job density (not a list of ZIP codes on your website). Years of continuous operation.

This data is the highest-value input because it answers the question the consumer is actually asking. When someone asks "who is the best HVAC company in Dallas," they mean: who does reliable work, consistently, in my area? Operational metrics answer that directly. Nothing else does.

This data also has maximum information gain because it exists nowhere else on the public web. It lives inside QuickBooks, ServiceTitan, Jobber, Dentrix, Clio, and the other systems businesses use to operate. No directory, no review site, no marketplace has it. For an AI system evaluating what new information a source provides, verified operational data is the most differentiating content possible.

Tier 2: Credentials, licensing, and insurance

State license number and status. Insurance coverage (general liability, workers' comp). Industry certifications. BBB accreditation and complaint history. Years of continuous licensing.

This data is verifiable against public sources (state licensing boards, insurance certificates of coverage) and provides the credentialing layer that separates legitimate operators from fly-by-night operations. AI systems can cross-reference it. It is not as differentiating as operational metrics, but it is far more structured than anything in Tier 3.

Tier 3: Reviews, business profiles, and website content

Google reviews. Yelp reviews. Google Business Profile. Website marketing copy. Directory listings.

This is what 98% of service businesses have and nothing else. The problem is not that it is bad data. The problem is that it is undifferentiating data. When every AI recommends a different plumber despite all of them having access to the same reviews, the reviews are clearly not sufficient for the AI to make a confident recommendation.

Reviews measure willingness to leave a review, filtered through recency bias and emotional state. A 4.8-star average across 200 reviews tells the AI that most customers had a good enough experience. It tells the AI nothing about job volume, service consistency, geographic coverage, or repeat customer behavior. When the AI has only Tier 3 data for every business in a category, it is guessing from fragments. Different AIs guess differently, which is why you get zero overlap across platforms.

For a deeper breakdown of exactly what data AI evaluates for each vertical, see the AI Data Priority Guides.

The gap: 98% of businesses are invisible

SOCi analyzed over 350,000 business locations and found that ChatGPT recommends 1.2% of them. Gemini recommends 11%. Perplexity, 7.4%.

For comparison, those same businesses appeared in Google's local 3-pack 35.9% of the time. AI visibility is up to 30 times harder to achieve than traditional local search ranking.

The reason is straightforward. Google ranks on proximity, reviews, and profile completeness. Those are low bars. AI systems need enough structured information to construct a confident recommendation. That is a high bar, and almost nobody clears it.

The gap is not going to close on its own. The businesses that are recommended today train the models that make recommendations tomorrow. Every citation reinforces the recommendation. Every business that never appears never gets that reinforcement. The advantage compounds.

What you can do about it

There are three layers to AI visibility for a service business, from simplest to most impactful. Each builds on the previous one.

1. Structured data on your website (Schema.org JSON-LD)

Check whether your website has LocalBusiness JSON-LD markup in the <head> of your pages. Not just name, address, phone number. Actual operational data: services offered, service area, years in operation, credentials.

Most local business websites either have no JSON-LD at all, or have broken markup auto-generated by a WordPress plugin that no one has audited. View your page source, search for application/ld+json, and read what is there. If the answer is "nothing" or "just NAP data," you have work to do.

The markup should use LocalBusiness or a more specific subtype (HVACBusiness, Plumber, Dentist, etc.). Never CreativeWork. Include every discrete, factual claim about your business that you can support.

This is table stakes. It does not differentiate you, but without it you are not even in the game.

2. An llms.txt file

An llms.txt file sits at your domain root (yourdomain.com/llms.txt) and acts as a navigation directory for AI crawlers. It tells the AI which pages on your site contain useful, structured information versus which pages are marketing copy. Some AI systems and indexing pipelines are beginning to check for it proactively.

We wrote a full guide on how to create one, with real examples. The short version: it is a plain Markdown file that links to your most data-rich pages with brief descriptions of what each contains. Do not list every page on your site. List only the pages where an AI can extract discrete, verifiable facts.

An llms.txt without structured data to point to is a map of blank territory. The file tells crawlers where to look. If there is nothing to find when they get there, the file adds no value.

3. Verified operational data via TrustRecord

The first two layers make your existing data more findable. This layer creates new data that does not exist anywhere else on the public web.

A TrustRecord is a structured, verified, machine-readable record of your business's operational history. It connects to the systems that run your operation (QuickBooks, ServiceTitan, Jobber, Dentrix, Clio), extracts real metrics through authenticated read-only connections, and publishes them in three layers: server-rendered HTML, Schema.org JSON-LD, and canonical JSON.

The business cannot edit, override, or selectively exclude any metric. That is the point. Independent verification is what gives the data credibility with AI systems, the same way backlinks gave pages credibility with Google.

Every TrustRecord is published in two places: on the business's own domain at /trustrecord, and on trustrecord.com as a canonical registry entry. Verified records are refreshed weekly from connected systems.

This is Tier 1 data. It is what moves a business from "one of 10,000 businesses with a 4.7-star rating" to "the business with 3,100 annual jobs, a 58% repeat customer rate, and 22 years of continuous operation, verified from authenticated accounting records." The AI can differentiate that. It cannot differentiate "quality service at affordable prices."

The compounding effect

There is a counterintuitive finding in Seer Interactive's data. Brands that are cited in AI Overviews earn 35% more organic clicks and 91% more paid clicks than brands that are not cited. Being a source the AI draws from does not reduce your traffic. It increases it. The AI Overview acts as an endorsement.

The same pattern holds for AI chat. Traffic from AI referrals converts at 4 to 23 times the rate of traditional search traffic. When a homeowner asks ChatGPT to recommend a plumber and then clicks through to the plumber's website, the decision is already mostly made. The AI did the comparison. The click is confirmation, not research.

Businesses with structured data get cited. Citations drive high-converting traffic. High-converting traffic generates positive outcomes. Positive outcomes reinforce the recommendation in future training data. The cycle compounds.

Businesses without structured data are not in the cycle. They are losing share to it.

The window

Right now, most local markets have zero businesses with verified, structured operational data published in AI-readable formats. The first business in a market to publish this data captures a disproportionate share of every AI-driven recommendation and comparison query in that category.

You cannot pay for AI recommendations. There is no AdWords for ChatGPT. The businesses that show up are the ones whose data makes them evaluable. And right now, in most markets, no one has done the work.

That will not last.

Deep dives by category

AI visibility looks different for every industry. The metrics that matter, the systems of record involved, and the competitive landscape vary significantly. These guides break it down by category, with vertical-specific analysis inside each:

• AI Visibility for Home Services Companies — HVAC, plumbing, roofing, electrical, pest control
• AI Visibility for Healthcare Practices — dental, medspas, medical practices
• AI Visibility for Professional Services Firms — law firms, accounting firms, insurance agencies