# Which Websites Does Google AI Cite for Professional Services? Lessons From 20 Canadian Studies

Source: https://enginely.ai/blog/which-websites-does-google-ai-cite-for-professional-services-lessons-from-20-canadian-studies/

> See which source types Google AI Overviews cite for professional services, based on 20 Enginely studies across four Canadian metros.

August 6, 2026

Google AI Overviews do not rely on one universal source type when answering professional-service questions. Sometimes the answer is built mainly from provider websites. In other markets, Google Business Profile listings, directories, regulators, public education sites, community forums and national chains take a meaningful share of the citations. The source pattern changes with the industry, the city and, most importantly, the question being asked.

That conclusion comes from Enginely's published series of 20 AI visibility studies covering dental care, legal services, physiotherapy, accounting and cosmetic surgery across the Greater Toronto Area, Ottawa, Calgary and Metro Vancouver. Each study used 21 high-intent queries across seven sub-markets, creating 420 city-level query runs in total. The reports captured the generated answer and classified every cited source by hand.

**The central finding: AI citation strategy cannot be reduced to 'publish more blogs' or 'get more reviews.' The source that wins depends on the evidence the question requires.**

## What the 20-Study Benchmark Actually Measures

The [published AI visibility benchmarks](https://enginely.ai/reports/) measure Google AI Overview citations for buyer questions such as treatment costs, professional fees, coverage, eligibility and service comparisons. They do not estimate impressions or assign an invented visibility score. They record the answer that appeared on a stated date, the sources attached to it and the businesses or organizations named.

The instrument is deliberately narrow. It covers Google AI Overviews, not ChatGPT, Perplexity, Gemini or Copilot. It is also a dated snapshot rather than a permanent ranking. Generated answers can vary between runs. The most useful findings are therefore structural: whether provider sites or listings dominate, how concentrated the citations are, which third-party sources repeatedly fill information gaps and how often citations cross local boundaries.

This distinction matters because a benchmark should guide decisions without pretending to reveal a fixed algorithm. A citation count can show where evidence is currently coming from. It cannot prove that one tactic caused the answer, and it cannot guarantee that the same domain will appear tomorrow.

## Five Source Patterns That Recur Across Professional-Service Searches

### 1. Provider Websites Win When They Publish the Decision Fact

The strongest own-site result in the published examples comes from dental searches in the Greater Toronto Area. Across 183 citations, 87 percent pointed to dental practice websites. The questions were not broad brand searches. They concerned implant pricing, Invisalign costs, routine cleaning, emergency fees and participation in the Canadian Dental Care Plan.

This pattern is commercially important. A practice homepage that promises quality care does not answer a cost or coverage question. A maintained page that states the relevant treatment, the pricing context, what changes the fee, the location and the date can. When providers publish usable decision information, Google has less reason to borrow the answer from a forum, insurer or aggregator.

The lesson is not that every provider needs to publish a fixed price. Many professional services cannot be priced responsibly without context. The lesson is to explain the pricing model, common ranges where supportable, inclusions, exclusions, variables and next step clearly enough that the page resolves the question.

### 2. Google Listings Can Become a Separate Citation Surface

Cosmetic surgery in the Greater Toronto Area produced a very different source mix. The study captured 294 citations, with 37 percent pointing to Google listings rather than clinic websites. Only 42 clinic websites received any citation, while 85 distinct Business Profile listings appeared. The top ten clinic websites held 55 percent of website citations.

This shows why a local profile should not be treated as a secondary copy of the website. In some answer sets, the listing itself becomes the cited object. Accurate categories, services, address details, practitioner information, reviews and current photos may therefore influence whether the business is represented at all. The listing and the website are different assets, and the market can use both.

The same tactic should not be copied blindly into every industry. Calgary accounting recorded zero Google Business Profile citations across 236 citations. For that market and query set, published pages carried the answer. Technical and content priorities should follow the measured source pattern, not a universal local SEO checklist.

### 3. Forums and Public Sources Fill Gaps Providers Leave Open

Ottawa accounting is the clearest example of a gap being filled by non-provider sources. Only 35 percent of citations pointed to an Ottawa firm. Twenty-four percent went to sources that were not accounting firms, including Reddit, the Canada Revenue Agency, national tax-preparation chains, Quora and lead-generation marketplaces. Reddit alone received 16 citations, more than every Ottawa accounting firm except one.

A forum does not need professional credentials to win a citation if it contains the specific fee discussion that provider sites avoid. The answer engine is trying to resolve the question with available evidence. When a profession publishes only generic service descriptions, public discussions can become the most detailed source in the accessible corpus.

This does not mean a firm should imitate an unverified forum comment. It means the firm should publish a better answer: clear assumptions, a date, scope, examples and professional review. The objective is not to silence the community source. It is to become the more reliable reference.

### 4. Directories Can Out-Cite the Businesses They List

In Metro Vancouver legal searches, a directory that published no street address received 13 citations, more than any law firm in the city. Reddit, the Law Society of British Columbia, public legal education sites and realtor websites also appeared repeatedly. Only 38 percent of citations reached a firm with a Metro Vancouver office.

Directories often win because they consolidate names, categories and comparison information on one page. A law firm may publish deep expertise but make it difficult to identify which lawyer handles which matter, where the service is offered or what the first stage costs. The directory may be less authoritative on the law but more usable for the exact shortlist question.

The response is not to manufacture directory listings or pay for indiscriminate inclusion. It is to improve authoritative profiles where they matter, correct inconsistencies and make the official site easier to use as the primary source.

### 5. Local Citations Do Not Always Respect City Boundaries

Several studies found businesses from other cities winning local questions. In the GTA dental report, 18 practices were cited outside the sub-market where they were located. Ottawa accounting answers sent 27 percent of citations to firms headquartered elsewhere. Calgary accounting included local firms that also appeared in Toronto and Ottawa studies.

Generated answers can treat geography differently from a map pack. A strong page about a service, fee or eligibility rule may be retrieved for a neighbouring city if it resolves the question better than local alternatives. For businesses, this creates both risk and opportunity. A nearby competitor can enter the answer without ranking in the local map results, while a strong source page can travel beyond its immediate market.

## A Cross-Study View of the Source Mix

**1. GTA Dental**

- **Source pattern:** 87% of citations came from practice websites.
- **What it suggests:** Detailed pages covering treatments, pricing, insurance coverage, and patient information are more likely to become primary citation sources.

**2. GTA Cosmetic Surgery**

- **Source pattern:** 37% of citations came from Google Business Profiles.
- **What it suggests:** Google Business Profiles can act as an important citation source alongside a business's website.

**3. Ottawa Accounting**

- **Source pattern:** 24% of citations came from non-provider websites, with Reddit cited 16 times.
- **What it suggests:** When accounting firms do not publish sufficient pricing or service information, AI systems may rely on forums and independent public sources instead.

**4. Calgary Accounting**

- **Source pattern:** No citations came from Google Business Profiles.
- **What it suggests:** Well-structured website content, rather than local business listings, provided the evidence used in AI-generated responses.

**5. Vancouver Legal**

- **Source pattern:** A legal directory received more citations than any individual law firm.
- **What it suggests:** Comprehensive comparison websites and directories can become the preferred reference when information across individual firm websites is incomplete or fragmented.

## What Determines the Source Type Google Uses?

The reports do not establish a universal causal formula, but they reveal a practical relationship between the question and the evidence available to answer it. The source type often reflects the job the answer must perform.

- Cost and fee questions favour pages that state numbers, ranges, inclusions and variables.
- Eligibility and coverage questions favour official programmes, insurers, government sources and provider pages that explain participation clearly.
- Shortlist questions can favour directories, listings and pages that connect a provider to a location and specialty.
- Professional-process questions can favour regulators, associations and public education sources.
- Local discovery questions may combine provider pages with Business Profile panels, depending on the vertical.

This is why keyword matching alone is an incomplete planning method. A content team should identify the evidence type needed for the decision, then decide which owned page, profile or credible third-party source should supply it.

## A Five-Layer Source Portfolio for Professional Services

Organizations should not try to win every answer from one page or one domain. A resilient source portfolio distributes accurate evidence across five layers.

### Layer 1: Official service and decision pages

Create one primary page for each distinct service and decision need. Include scope, process, eligibility, limitations, location relevance, evidence and a clear update owner. These pages should be the most complete official explanation of the organization.

### Layer 2: Practitioner and organization entities

Connect the organization, professionals, credentials, services and locations consistently. The relationship should be visible in page copy, navigation, structured data and approved external profiles.

### Layer 3: Local business data

Maintain the Business Profile and local listings as independent data assets. The site may dominate one market while listings dominate another. Hours, categories, services, addresses, appointment links and review responses should remain current.

### Layer 4: Professional and third-party corroboration

Regulator profiles, associations, respected directories, media coverage and research references help confirm claims the organization cannot validate by repeating them on its own site. Pursue accurate inclusion, not artificial mentions.

### Layer 5: Public-question monitoring

Forums and community sources reveal the questions formal websites are failing to answer. Use them as research inputs, then publish a more accurate and accountable response through an expert-reviewed process.

Enginely's [industry-focused GEO programmes](https://enginely.ai/industries/) adapt this source portfolio to the evidence, professional rules and buying questions that differ across dental, legal, healthcare and other high-trust markets.

## How to Run a Source Benchmark for Your Own Market

A useful benchmark begins with the customer decision, not the company name. The following method produces evidence that can guide implementation.

1. Choose three to five commercially meaningful question groups, such as fees, eligibility, comparisons, timing or provider selection.
1. Create natural variations for each question and define the cities or sub-markets that matter.
1. Capture the complete answer, the businesses named and every cited URL on a stated date.
1. Classify each source as an owned website, listing, directory, regulator, government source, forum, publisher, marketplace or other category.
1. Verify provider locations and credentials manually before calling a source local or professional.
1. Separate presence, citation, accuracy and commercial action. A cited page is not automatically a lead.
1. Repeat the benchmark on a consistent schedule and report structural movement rather than one favourable answer.

A specialist [AI citation optimisation service](https://enginely.ai/services/generative-engine-optimization/) should use the benchmark to identify the failing layer: missing decision content, weak technical access, fragmented entity information, incomplete local data or absent corroboration.

## What Businesses Should Do With the Findings

The most valuable next step is not to copy the current winner. It is to understand why that source was usable for the question and build a more accurate answer within the organization's control.

- If provider websites dominate, improve the specific pages that answer costs, process, suitability and eligibility.
- If listings dominate, correct local data and connect the profile to strong service and practitioner pages.
- If directories dominate, strengthen official pages and repair important professional listings.
- If forums dominate, identify the missing facts and publish a reviewed answer with assumptions and dates.
- If out-of-market providers dominate, build stronger location-service evidence instead of producing thin city pages.
- If the market is highly concentrated, analyse the leading sources before deciding whether the opportunity is editorial, local, technical or authority-led.

For dental practices ready to act on a defined report, Enginely's [fixed-scope AI Visibility Accelerator](https://enginely.ai/services/ai-visibility-accelerator/) connects profile, technical, content, authority and remeasurement work within a stated programme.

## What This Benchmark Does Not Prove

The studies are not a list of ranking factors. They do not show that a directory was cited because it is a directory, or that a listing caused a clinic to be selected. They record what appeared, not every internal decision that produced the answer.

They also cover one Google surface at one period in time. Other platforms have different crawlers, retrieval systems, product contexts and citation behaviour. The correct use of the findings is to identify source patterns worth testing, not to promise that one content format will work everywhere.

## Frequently Asked Questions

### Do provider websites or directories receive more AI citations?

It depends on the market and question. GTA dental citations overwhelmingly favoured practice websites, while Vancouver legal included a directory that out-cited every local firm. A benchmark must classify the actual sources used in the relevant market.

### Are Google Business Profiles important for AI Overviews?

They can be. GTA cosmetic surgery assigned 37 percent of citations to Google listings, while Calgary accounting recorded none. The importance of the listing varies by vertical, query and location.

### Should a business publish exact prices to get cited?

Only when exact prices are accurate and appropriate. Many professional services require ranges, assumptions and variables instead. The useful objective is to answer the pricing question responsibly, not to oversimplify it.

### Does an AI citation guarantee an enquiry?

No. A citation measures source use inside an answer. It may influence awareness or a shortlist without producing a trackable click. Commercial reporting should also examine referral traffic, branded demand, enquiries and lead quality.

## Build the Source the Question Actually Needs

The 20 Canadian studies show that AI visibility is a source-design problem as much as a content problem. A provider website wins when it publishes the decision evidence. A listing wins when local data becomes the answer. A directory or forum wins when providers leave a comparison or fee question unresolved.

The practical strategy is to map each important customer question to the source best qualified to answer it, then make that source accessible, specific, accurate and maintained. The goal is not to control the generated answer. It is to ensure that credible evidence exists wherever the answer system looks.
