Local SEO in the Age of AI Overviews and ChatGPT

How service businesses get cited

The answer box now sits above the map. Someone in Coquitlam searches "furnace repair near me," and before they reach the local pack, an AI Overview has already summarized who to call and why. Ask ChatGPT the same question and it returns three or four businesses by name, with links.

Short answer: No one sells a slot in an AI answer. The engines assemble their recommendations from your Google Business Profile, your reviews, and your web pages. Make your business unmistakable to a machine and you win the local pack and the AI answer at the same time.

Free. No pitch for a service you don't need.

Being one of those named businesses is the new front row.

And here's the part worth saying plainly: the engines are not inventing those recommendations. They assemble them from your Google Business Profile, your reviews, and your web pages, plus a few signals that are new. No magic. No guaranteed slot. Just a different set of inputs feeding a different kind of answer.

We do not guess about which inputs matter, so this piece separates what the research actually shows from what agencies are currently selling. Let's start with a word that trips people up.

Terms

What "cited" actually means here

Cited, in this article, means your business named or linked as a source inside an AI-generated answer. When ChatGPT lists your company and links to your page, that’s a citation. When an AI Overview summarizes your service and links back to your site, that’s a citation.

That is a different thing from a local citation in the older sense. A local citation is a listing of your name, address, and phone number on a directory like Yelp, YP.ca, or an industry association. Those still matter, and we’ll come back to them, because consistent listings are one of the ways an AI engine figures out who you are.

AI citation — your business named or linked inside a generated answer.
Local citation — your name, address, and phone number listed on a directory.
Below, “get cited” always means the AI answer, not the directory listing.

Keep the two straight and the rest of this reads clearly.

01 — The timeline

What changed, and when

AI Overviews arrived faster than most local businesses noticed. These are Google's own dates, not an estimate.

May 2024

AI Overviews launch in the US at Google I/O

Oct 2024

The feature reaches Canada

May 2025

200+ countries and territories, 40+ languages

Aug 2025

AI Mode arrives in Canada, 21 August

AI Mode is the full conversational tab, where the entire results page is generated rather than listed. Google has also begun placing ads inside AI Overviews, which tells you where the company expects attention to land.

For a service business in Langley or anywhere across Metro Vancouver, the consequence is concrete: more searches now resolve before the click. A homeowner used to scan the local pack and pick. Now a summary often makes the shortlist for them, and your job is to be on it.

02 — The pipelines

How the engines actually pick who to name

Two engines dominate local questions right now, and they run different pipelines. Knowing the difference tells you where to put your effort.

Google

AI Overviews and AI Mode

They lean on what Google already ranks. For a local query, that means the local pack, the Google Business Profiles behind it, and the reviews attached to them.

The AI layer reads those signals, summarizes, and links. If you already rank in the three-pack for “electrician Burnaby,” you are in the pool the Overview draws from. See how we run Local SEO →

OpenAI

ChatGPT

It works differently. Its browsing mode pulls live results from Bing’s search index, not Google’s, and typically returns three to six numbered citations per answer, according to analysis published by ZipTie in March 2026.

It weighs domain trust, content quality and freshness, and whether your business shows up consistently across the platforms it can see. Reviews on third-party sites feed that read. See how we run Answer Engine Optimization →

One shared step sits underneath both. The engine has to resolve you to an entity before it can name you. It needs confidence that your business, the phone number on your website, the profile on Google, and the listing on a directory are all the same company. A consistent name, address, and phone number — your NAP — is how it gets that confidence. Entity clarity is not a nice-to-have here. It is the precondition for being cited at all.

03 — The evidence

What the evidence says actually gets cited

The strongest evidence we have comes from a peer-reviewed source, not an agency blog.

Researchers from Princeton and collaborators published “Generative Engine Optimization” (Aggarwal et al.), presented at KDD 2024 and tested across a 10,000-query benchmark they built called GEO-Bench. They measured which content changes moved a page’s visibility inside AI answers. Three moves worked. One did nothing.

+30–40%

visibility lift from adding citations to credible sources

+40%

from adding relevant statistics

+41%

from adding quotations from credible sources

0%

from keyword stuffing: no benefit for generative engines

Writing with sources, numbers, and quotes is what earned the citation. The old habit of repeating a phrase to rank produced nothing.

Structure compounds those gains. Independent analyses of ChatGPT citations found that a large share, around 44 percent, come from the first third of a page, which means a front-loaded answer beats a buried one. Pages that keep a clean H1 to H2 to H3 heading hierarchy get read more reliably. Freshness helps too: recently updated pages tend to be cited more often than stale ones.

One more finding is worth your attention because it breaks with old SEO instinct. An Ahrefs study of tens of thousands of brands in 2025 found that brand mentions across the web correlated with AI visibility more strongly than backlinks did. Being talked about, by name, on pages the engine trusts appears to matter more than the link-building most agencies still lead with.

A caution, in keeping with how we read data. Some of the numbers circulating about AI search come from single vendors reprocessing their own samples, and they should be treated as directional, not settled. The Princeton study, Google’s own launch dates, and the Ahrefs sample are the load-bearing sources here. The tidy single-source percentages you’ll see quoted elsewhere are worth a raised eyebrow.

04 — A worked example

A Metro Vancouver worked example

This is an illustration, not a named client, but every step maps to a real signal.

Picture a furnace-repair company in Coquitlam. A homeowner two neighbourhoods over searches “furnace not working Coquitlam” on their phone in January. Google returns an AI Overview: a two-sentence summary of what to check, then a short list of local companies to call, each linked.

The same homeowner, unsure, opens ChatGPT and asks, “who can fix a furnace fast in Coquitlam?” ChatGPT returns four businesses with links pulled from its Bing-indexed view of the web. Three companies could have been named. One was, in both answers. Walk back why.

 The one that got namedThe one that didn't
Google Business ProfileCategory "HVAC contractor," service area set, hours completeHalf-filled, no service area
ReviewsForty, a dozen from the last ninety daysA thin, ageing pile
NAP consistencyIdentical on site, profile, and two directoriesMismatched across listings
The page it landed on"Emergency Furnace Repair in Coquitlam," answered up front, updated that winterA homepage saying "Your Comfort Is Our Priority"
ResultNamed in both answersNamed in neither

The winner’s NAP matched character for character across all three places it appears, so both engines resolved it to a single, confident entity. Its Coquitlam page cited the local permit requirement instead of gesturing at comfort. Nothing about that second homepage told either engine what it did, where, or for whom. The engine could not cite what it could not resolve. The example points at specific, checkable moves. Here they are, in order of leverage.

Not sure what an AI engine can currently resolve about you?

We'll run the queries your customers actually type and show you who gets named instead.

Book a free audit
05 — The playbook

What a service business should actually do

Work this list top-down. The earlier items resolve you to an entity; the later ones make you the entity worth naming.

Finish the Google Business Profile

Set the primary category precisely, add secondary categories that match your real services, define the service area, complete hours and attributes, and keep photos and Q&A current. This profile is the entity Google's AI reads first, and secondary categories change which queries you surface for.

Make your NAP identical everywhere

Your name, address, and phone number should match, character for character, on your site, your Google profile, and every directory. This is not vanity. It is how an engine confirms the listings all point to one business. Mismatched suite numbers and old phone numbers quietly split your entity in two.

Build reviews with recency, and respond

A wall of reviews from two years ago reads as a business that stopped. Recent reviews, and owner responses to them, give both engines a trust signal they can actually read. This is the same trust step from the pipelines above, now something you act on.

Front-load the answer on real questions

Take the questions customers actually type, give each its own page or section, and answer in the first paragraph. Add a relevant statistic, cite a real source, and put a visible updated date on it. That is exactly the combination the Princeton study found moved citations.

Earn third-party mentions

Get named by local press, community associations, supplier pages, and reputable directories. Mentions, by name, on trusted pages track with AI visibility more closely than raw link count, so a feature in a local outlet can outperform a dozen low-quality links.

Add structured data

Mark up your pages with schema (LocalBusiness, with your service area and services) so the machine reads your attributes without having to infer them. It reduces the guesswork between your page and the engine's model of your business.

None of these are new inventions. Every one of them also strengthens your local pack ranking, which is the point: the durable work pays off in the map and in the AI answer at the same time.

06 — The limits

What we don't know yet

Honesty about the gaps is part of doing this well, so here are the real limits.

Volatile: the same query can name different businesses from one session to the next
Unguaranteeable: no agency can promise an AI citation, us included
Directional: much of the published data comes from vendors measuring their own tools
Not a click: being cited is not the same as being visited

Studies have found that only a minority of brands hold a consistent AI presence across sessions. Anyone promising you a permanent spot in an AI Overview is selling something the technology does not currently offer. We can make your business far more citable, by resolving your entity, strengthening reviews, and structuring content the way the evidence supports. We cannot promise a model will name you on a given day, and you should be wary of anyone who does.

And being cited is not the same as being clicked. AI answers can satisfy a searcher without sending them to your site, which changes what you measure. The businesses that do well treat the AI answer as the new shelf position, and the phone call or booking as the result that still counts.

The through-line is the one we opened with. Do the durable work, finish the profile, earn the reviews, make your entity unmistakable, and publish genuinely useful pages, and you win the local pack and the AI answer together. We do not guess about that, because the evidence and the mechanics point the same way.

07 — Answers

Frequently asked questions

ChatGPT’s browsing mode pulls live web results from Bing’s search index, not Google’s, according to analysis published in March 2026. That’s one reason your presence on Bing Places and across third-party platforms matters for ChatGPT visibility, even though Google still drives AI Overviews.

AI Overviews reached Canada in October 2024, after launching in the United States at Google I/O in May 2024. Google’s fuller AI Mode arrived in Canada on 21 August 2025.

Generative Engine Optimization overlaps heavily with SEO but is not identical. GEO targets being named inside AI-generated answers, and the Princeton GEO study found it rewards citations, statistics, and quotations rather than keyword repetition. The strongest local SEO foundations, a complete profile and consistent entity signals, feed both.

Reviews influence which local businesses AI engines surface, because both Google’s and ChatGPT’s pipelines read review signals as evidence of trust. Recency and owner responses strengthen that signal, and a steady flow of recent reviews reads more favourably than an older, larger pile.

AI Overviews can reduce clicks by answering a question in the results page, so some businesses see fewer visits even when they’re cited. The response is to treat the AI answer as a placement worth winning and to measure calls and bookings, not only page views.

No honest provider can guarantee an AI citation, because the engines vary their answers session to session and offer no paid or permanent placement in the organic AI summary. What’s controllable is making your business more citable: resolving your entity, building reviews, and structuring content the way the evidence supports.

Ready to see what the engines can find about you?

We run an audit that checks exactly what an AI engine can currently resolve about your business, where your entity signals break, and which pages are and aren't citable. Book a free audit and we'll show you the gaps, and the durable work that closes them.

Book a free audit

Quarc Digital: Local SEO & Digital Marketing Scientists · Langley, serving Metro Vancouver & the Fraser Valley · (778) 726-1420