I asked ChatGPT to recommend a plumber in Vancouver. Here is what came back, and what the named businesses all had in common.

A first-person experiment. Same prompt across ChatGPT, Perplexity, and Google AI Overviews. The businesses that surfaced were not random. They shared five very specific traits, and any service business can replicate them.

A few mornings ago I decided to run a small experiment. Nothing scientific. Just the same question, posed three times, to the three AI surfaces quietly rewriting how customers find local services. I opened ChatGPT, then Perplexity, then Google AI Overviews. Into each one, I typed the same prompt: "Best emergency plumber in Vancouver BC."

I cannot, in good faith, list the specific names the models surfaced. Those answers shift from session to session, and naming them here would freeze a moving target. But the more interesting finding was not which businesses got cited. It was the pattern. The handful of names returned by each surface shared a strikingly consistent profile. And the truth is, this profile is reproducible. Any service business willing to do the work can build it.

The experiment, in plain terms

ChatGPT gave me a short list. A handful of names, each accompanied by a sentence or two of justification. Perplexity returned a similar count, but with linked citations underneath: Yelp reviews, BBB profiles, a couple of local news mentions, a HomeStars listing. Google AI Overviews was the most compact. Two or three names, each tied to a Google Business Profile, with a short rationale pulled from review content.

The names overlapped, but not entirely. ChatGPT and Perplexity agreed on two businesses. Google AI Overviews surfaced a slightly different set, though one name appeared on all three.

What stood out was this. The cited businesses were not the ones I would have predicted from a traditional Google search. The biggest paid-ad spenders did not appear. The companies with the slickest websites did not always appear. The ones that did appear, however, shared a profile so consistent it almost felt like the models were following a checklist.

Trait one: a Google Business Profile that breathes

Every business named across the three surfaces had a Google Business Profile that was, to put it simply, alive. Photos uploaded recently. Q&A section populated with real questions, real answers. Posts in the last few weeks. Reviews accumulating at a steady, ongoing pace rather than a single burst followed by silence.

This last point is worth dwelling on. AI systems do not just look at total review count, nor only at star rating. They appear sensitive to review velocity, the rhythm of new reviews arriving over time. A business with two hundred reviews, all from three years ago, reads as stale. A business with eighty reviews, but with new ones arriving every couple of weeks, reads as a going concern.

It must be said: this is not gamed easily. You cannot fake a heartbeat. You have to actually be asking customers for reviews, consistently, as part of the workflow.

Trait two: structured data that tells the AI what you are

Every cited business, when I pulled up their website source, had LocalBusiness schema. Most had Service schema declaring the specific services offered. Some had FAQPage schema on their question pages. A few had Person schema for individual technicians.

This is not glamorous work. Structured data lives in the page's <head> as JSON-LD, invisible to the casual visitor. But to an AI crawler constructing an answer, it is the difference between reading a marketing brochure and reading a labelled specification sheet. Naturally, the labelled sheet is easier to cite. Our earlier piece on why your website may be invisible to ChatGPT and Perplexity covers the JSON-LD details, including the exact LocalBusiness fields each surface seems to read first.

Trait three: FAQ pages that answer real, specific scenarios

Here is where it became almost comically obvious. The named businesses all had, without exception, pages dedicated to specific plumbing scenarios. Not "Our Services" pages. Not "About Us" pages. Pages titled like "What to do when your hot water tank starts leaking" or "How to shut off the main water valve in a Vancouver house" or "Is a slow drain an emergency or can it wait until Monday?"

Each page answered the question directly, in the first paragraph. Then expanded with detail. Then tied the answer back to the business's services without being aggressive about it.

This is the format AI systems were built to consume. ChatGPT and Perplexity do not want to parse your hero banner and three rotating testimonial slides. They want a page that says: here is the question, here is the answer, here is who is qualified to help.

Trait four: citations beyond Google

This was perhaps the most overlooked pattern. The businesses that surfaced did not exist only inside Google's ecosystem. Their names appeared consistently across third-party directories: Yelp, BBB, Angi, HomeStars, and in a couple of cases local trade-association or BNI directories.

Why does this matter? Because AI models do not source their information from a single place. ChatGPT draws on Bing's index and a mix of trusted web sources. Perplexity searches broadly. Google's AI surfaces lean on Google's own data, but the corroboration provided by external citations strengthens the signal.

A business that exists only on its own website and its GBP is a single data point. A business that appears consistently across eight or ten independent sources, with matching name, address, and phone number, becomes a verified entity. Of course, this overlaps with the directory work covered in our local SEO checklist for 2026.

Trait five: named humans, not faceless brands

This one surprised me the most. Every cited business, every single one, had a website where named technicians appeared somewhere. A team page with photos and short bios. A founder's letter. A blog post signed by a specific person. Not "the team at [company]" but "Marco, our master plumber with 14 years of experience."

A few went further and marked up these people with Person schema, declaring job title, experience, and certifications. This is E-E-A-T made machine-readable. The AI is not just reading "we have experienced plumbers"; it is parsing a structured claim that a specific named person performs this work.

Faceless brand websites, the ones with stock photos of smiling actors and no actual humans named anywhere, did not surface in my results. Not once. The lesson, without doubt, is that AI systems are quietly rewarding the companies willing to put real people forward.

Why these five traits cluster together

It is tempting to treat each item as a separate task to tick off. That would be a mistake. The five traits cluster together because they are all expressions of the same underlying condition: the business is a genuine, ongoing operation with verifiable people doing verifiable work, and it has made that legibility easy for machines.

A business with strong reviews but no schema sends mixed signals. A business with rich schema but no third-party citations makes claims that cannot be corroborated. The AI is not looking for any one trait. It is looking for the constellation. For the longer playbook on engineering this deliberately, our piece on how to get your business mentioned by ChatGPT and Perplexity breaks down the workflow step by step.

You can run this same test in sixty seconds

You do not need to trust my Vancouver plumber experiment. You can run the equivalent test for your own city, your own trade, right now.

Open ChatGPT. Type: "Best [your service] in [your city]." See what comes back. Open Perplexity. Ask the same thing. Open Google and look at the AI Overview at the top of the results. Then ask yourself: did my business appear? If not, which businesses did appear, and what do they have in common?

Pull up their websites. View source. Look for LocalBusiness schema, Person schema, FAQ pages on specific scenarios. Open their GBP and check recent activity. Search their name across Yelp, BBB, and your industry's main directory. You will find the pattern. It is the same pattern, in city after city, trade after trade.

This is the new local SEO, except it is not really SEO anymore. Call it reputation engineering for machines. The window of competitive advantage here is narrow, and closing. None of this is gated behind enormous budgets; it is gated behind attention and follow-through. For a more granular treatment of the underlying ranking mechanics, our guide on ranking inside ChatGPT in 2026 goes deeper.


Run the same check on your own site. Paste your domain into licheo.com/get-found-check. Sixty seconds, no email required.

Frequently asked questions

Does ChatGPT actually recommend specific local businesses by name?

Yes. For service queries in well-populated metropolitan areas, ChatGPT frequently surfaces specific business names with a short justification for each. The behaviour is more reliable when the prompt is specific, such as "emergency plumber in Vancouver BC" rather than just "plumber". Smaller towns and highly niche services produce more generic responses, though the model still names businesses where the underlying data supports it.

What do the businesses ChatGPT recommends have in common?

Five traits appeared in every business named across ChatGPT, Perplexity, and Google AI Overviews: a Google Business Profile with recent activity, LocalBusiness schema on the website, pages answering specific customer scenarios, consistent listings across third-party directories, and named human staff rather than a faceless brand. No cited business was missing any of the five.

What kind of pages get a service business cited by AI?

Pages built around one specific customer scenario, not service overviews. The cited plumbers had pages titled like "What to do when your hot water tank starts leaking" or "Is a slow drain an emergency or can it wait until Monday?". Each answered the question directly in the first paragraph, expanded with detail, then connected the answer to the business's services without pressure.

Why do third-party directory listings matter for AI recommendations?

Because AI models corroborate across sources rather than trusting one. Every business that surfaced appeared consistently on Yelp, BBB, Angi, HomeStars, and in some cases trade-association directories, with matching name, address, and phone number. A business existing only on its own site and its Google profile is a single data point; one appearing across eight or ten independent sources reads as a verified entity.

Does naming your staff on your website help with AI search?

It appears to matter a great deal. Every cited business had named technicians somewhere on the site: team pages with photos and bios, a founder's letter, or posts signed by a specific person. Not "the team at [company]" but "Marco, our master plumber with 14 years of experience". Faceless sites with stock photos and no named humans did not surface once.

How often do ChatGPT's business recommendations change?

More than most people expect. Running the same prompt three times in one afternoon can produce partial rotation, particularly among the lower-ranked names. The top one or two businesses tend to stay stable, because their signal is dense and corroborated enough that the model returns them consistently.

Can I pay to appear in ChatGPT recommendations?

No. There is no advertising product inside ChatGPT or Perplexity for local recommendations today. The only levers are the five traits: an active Google Business Profile, structured data, scenario-specific pages, distributed third-party citations, and named humans. Agencies promising paid "ChatGPT optimisation" have nothing concrete behind the offer.

How long does it take to get recommended by AI search?

Realistically three to nine months of consistent work. No single element takes that long: schema markup and named-team pages can be deployed in a week. What cannot be compressed is review velocity and the accumulation of third-party citations, which depend on sustained operational discipline rather than a one-time project.

How can I test whether AI assistants recommend my business?

It takes about a minute. Ask ChatGPT for the best provider of your service in your city, ask Perplexity the same question, then check the AI Overview on the same Google search. Note which businesses appear. Open their sites and view the source for LocalBusiness and Person schema, check their Google profiles for recent activity, and look them up across the main directories in your trade.

Do these AI citation patterns apply to industries other than plumbing?

Yes, the mechanics translate cleanly and only the directory names change. A lawyer's third-party citation set is Avvo, Martindale-Hubbell, and state bar listings. A dentist's is Healthgrades, Zocdoc, and college directories. A restaurant's is OpenTable, Tripadvisor, and local food blogs. Profile activity, schema, scenario pages, distributed citations, and named humans apply across any service industry.

Rather not do this yourself?

We can simply do it for you

Everything in this article, the website fixes, the content, being found on Google and inside AI assistants like ChatGPT, is work that big companies pay a whole team of specialists for. Licheo does it for small businesses, every month. You never learn a tool, and you are never handed a to-do list. See what that team costs a big company.