AI Pulse: AI Search Clicks Go Local — Most Canadian Practices Aren't Capturing Them

Week of 2026-05-04 · Manta AEO

A new Similarweb analysis of ten markets confirmed this week what local practice owners should act on: when people click from AI-generated answers, they click local. The structural advantage belongs to practices with strong local signals. Most haven’t built them yet, and the gap between practices that show up and practices that don’t is widening pretty much every week now.

This Week in AI Search

AI Search Clicks Favour Local Domains, Not National Brands

Analyst Aleyda Solis published a Similarweb-based analysis covering AI-generated click patterns across ten markets. The finding is direct: clicks from AI search results consistently go to local domain versions, not national or global brand sites. The distribution varies by industry, but the pattern holds across markets. For an independent chiropractor in Mississauga or a solo accountant in Halifax, this is a structural advantage — provided their local signals are in place. Consistent NAP data, a verified Google Business Profile, and content that names a specific city are how AI platforms decide “this result is local.” Practices missing any of those three are competing nationally when they should be winning locally. AI Search Clicks Often Go To Local Domains: Report

Researchers Benchmark 11,500 Queries to Map How AI Overviews Select Sources

A new academic paper published this week introduced a benchmark dataset of 11,500 user queries examining how Google Search, Gemini, and AI Overviews retrieve and present information differently from traditional search. The key finding: the sources appearing in AI Overviews are not always the sites ranking first organically. AI Overviews apply their own source selection logic — separate from the ranking algorithm. For an independent optometrist or physiotherapy practice in Ontario, this means a Google ranking and an AI citation rate are two different scores that need separate work. Building one does not automatically build the other. How Generative AI Disrupts Search: An Empirical Study

AI Visibility Tracking Tools Are Producing Unreliable Numbers

Search Engine Journal flagged a growing measurement problem this week: many of the tools businesses use to track AI visibility are introducing noise that makes it difficult to separate real citation gains from artificial signals. The specific failure mode is measurement tools that inflate visibility numbers — showing rising scores that don’t reflect what a patient or client actually sees when they search. For any practice owner who has started monitoring their AI search presence, the implication is clear: the number a tool displays may not match reality. Direct prompt testing — asking ChatGPT or Perplexity the question your clients would ask, and reading the response — remains more reliable than dashboard metrics. Your AI Visibility Tracker Is Quietly Breaking Your Analytics

AI Platforms Reward Structured Comparison Content

HubSpot published analysis this week on how AI platforms decide which businesses to surface when a query implies comparison — “best dentist in Oakville,” “who should I call for estate planning in Kitchener.” The mechanism: AI platforms read structured content that names the competitive context, explains what makes a practice different, and answers comparison questions directly. A practice with no FAQ section and no “why choose us” content is invisible in this answer type. These comparison queries are high-intent — the person asking has already decided to act and is choosing between options. AEO Competitor Analysis: Track AI Answer Engine Rivals

New Data on What’s Getting Cited

Local Signals Drive Local AI Citation — Across Ten Markets

The Similarweb analysis covered industry-level breakdowns across ten markets. The pattern is consistent: AI platforms apply local weighting when selecting sources. A dental practice in Burlington with a complete Google Business Profile, consistent address data across directories, and content that mentions Burlington specifically is a stronger AI citation candidate than a practice with better organic rankings but weaker local infrastructure. The gap between ranked-but-not-cited and ranked-and-cited increasingly comes down to local signals, not content volume. Practices in mid-sized Ontario cities — Barrie, Guelph, Kingston, St. Catharines — face less local competition for those AI citations than practices in Toronto. The advantage is available. AI Search Clicks Often Go To Local Domains: Report

Google Rankings and AI Citations Are Diverging

The 11,500-query benchmark documents cases where highly-ranked pages are not cited by AI Overviews, and where lower-ranked pages are cited because of structural content characteristics. The researchers track how source selection differs between traditional organic results and AI-generated answers. For any independent practice that has spent years building Google rankings: those rankings still matter, but they no longer transfer automatically into AI citations. An independent financial advisor in Waterloo who ranks on page one for “IIROC advisor Waterloo” may still have zero AI Overviews citations if the site’s content structure doesn’t answer the specific questions AI systems are scanning for. How Generative AI Disrupts Search: An Empirical Study

Measurement Noise Is a Real Problem, Not a Vendor Issue

The Search Engine Journal analysis goes beyond individual tool criticism. The core problem is structural: many AI visibility metrics are proxies, not direct measurements. When a tool reports your “AI visibility score” improving, it may be measuring a signal that correlates with citations rather than citations themselves. This matters most for practices evaluating whether their current strategy is working. Manta audits citations through direct prompt sampling — asking the platforms the questions your clients ask and recording what they say — rather than relying on proxy metrics. That approach is slower to produce data but produces data that reflects what patients and clients actually see. Your AI Visibility Tracker Is Quietly Breaking Your Analytics

What This Means for Independent Canadian Practices

The local domain finding is the most immediately useful piece of data this week for Ontario independent professionals. AI platforms apply local weighting, and practices with consistent local signals win that weighting. An independent physiotherapist in Barrie who has their clinic’s address, neighbourhood, and a few city-specific FAQ answers on their site is a better AI citation candidate than a better-resourced Toronto chain with stronger domain authority. Geography is an advantage here, not a limitation. Practices in cities where local competition is thin — Sudbury, Sault Ste. Marie, Peterborough, Lethbridge — have an even clearer path to local AI citations than their counterparts in crowded urban markets.

The divergence between Google rankings and AI citations is a conversation every practice owner should have with whoever manages their website. If an agency is reporting keyword rankings and calling it AI visibility, those are different things and the second is the one changing fastest. Ranking first for “chiropractor Oakville” and appearing in the AI answer to “who’s the best chiropractor in Oakville” require overlapping but distinct work. Understanding the gap between the two numbers is the starting point.

The measurement reliability issue is a quiet problem. Any practice evaluating an AI visibility strategy based on a dashboard score should cross-check that number with a direct test: open ChatGPT and Perplexity, type the question a client would ask, and read the response. What the tool says and what the platform actually says may not match.

One Thing to Do This Week

Open your Google Business Profile and check three things: your city name appears in the business description, your specialty is named explicitly, and at least one FAQ answer is filled in. Then open ChatGPT and type “best [your profession] in [your city].” Compare what appears to what your profile says. If you’re not in the results, the profile description is the fastest local signal to fix — it takes ten minutes and no developer. That description is one of the first local signals AI platforms read when deciding which practices to cite.

Sources

Last updated: 2026-05-04