How AI Search Decides Which Ontario Massage Therapists to Recommend

Here is the claim most massage-marketing advice gets backwards: the College of Massage Therapists of Ontario, the regulator that tells you what you may and may not advertise, has more influence over whether an AI engine recommends your practice than your website designer does. That sounds like a stretch until you follow the mechanism. AI engines build a recommendation by reading factual, attributable third-party pages. CMTO advertising rules force every page you control to be factual and attributable, and strip out the testimonials and superlatives other industries use to win. The regulator, in effect, has been writing GEO-friendly copy rules for years without calling them that.

So when an Ontario resident opens ChatGPT and types "best registered massage therapist near me," the answer is not assembled from the practice sites the engine crawled. It comes from a short stack of sources the engine could fetch and trust in that moment, mostly directories and listing pages, filtered through what the rules let anyone say about a regulated health provider. Understanding that stack is the whole game. This piece walks through what the engines actually read for this vertical, why the CMTO frame changes the strategy, and what a single-location clinic can do about it.

How an AI engine actually builds a "best RMT near me" answer

An AI engine answers a local health query by running a retrieval step, pulling a handful of pages it can fetch and trust, reading the names inside them, and writing a short list. It does not open the web, weigh every clinic, and rank them the way Google's classic ten blue links did. It grabs a few citable sources and synthesises. A directory that already lists ten registered massage therapists with addresses and a one-line strength each is far easier to ground than a polished clinic homepage written as reassurance for a first-time client. So the directory earns the citation, and the practice site sits unread.

The retrieval-and-synthesis behaviour that decides which Ontario clinic gets named already has a formal name in the literature. Pranjal Aggarwal and colleagues described it in "GEO: Generative Engine Optimization" (arXiv:2311.09735), presented at KDD 2024. Generative engines answer by synthesising and summarising several cited sources rather than returning a ranked list of links, and the paper measured that adding citations, quotations, and statistics to a source can raise its visibility inside AI answers by up to 40 percent. Being a citable source is the lever. Holding position one on Google is not. A Guelph clinic can rank first for "massage therapy Guelph" and still lose the AI answer to a directory page the engine found easier to read.

Read that carefully and a strategic shift falls out. You are not trying to out-design competitors for a human's eye. The real job is showing up, accurately and extractably, in the exact registries and directories each engine pulls into its retrieval step. For a regulated health service in Ontario, that source list has a very particular shape, and it starts somewhere no other vertical starts.

The CMTO register is the source every honest answer should trace to

The College of Massage Therapists of Ontario runs a free, always-current public register called Find an RMT, and it is the single most authoritative source about who is actually a registered massage therapist in the province. Anyone can search it by name, practice location, or postal code, and confirm registration status, practice locations, languages spoken, and conduct history. CMTO's own "Look Before You Book" campaign exists to push the public to check that register before they book. In entity terms, this is your ground truth: the record that proves your practice exists as a regulated health provider, the way the machine wants existence proven.

Most RMTs have never thought of the register as a marketing asset, because it is not. It is a regulatory database. But AI engines treat regulator-run registries as high-trust sources, precisely because they are neutral, verifiable, and free of promotional language. When an engine is asked to name a legitimate RMT and wants to avoid recommending an unregulated masseur, a College register is exactly the kind of source it reaches for to disambiguate. Your register entry being complete and correct is not optional hygiene. It is the anchor the rest of your AI visibility hangs from. In the 12 Vectors, this is Vector 2, Anchor: establishing an unambiguous, verifiable entity the retrieval systems can attach every other signal to.

In Matt's audits of Ontario registered massage therapists, the pattern is consistent and a little painful: the clinic has spent thousands on a website and almost no attention on the free, authoritative sources the engines actually read. The register entry has a stale address. The Google Business Profile lists the wrong hours. Two directories spell the practice name three different ways. None of it shows on the pretty homepage, and all of it is exactly what an engine trips over when it tries to decide whether you are a real, consistent entity worth naming.

Which directories and listings AI reads for Ontario RMTs

A short roster of platforms supplies most of the sources an engine can reach for this vertical, and those names are the practice owner's real battleground. If your clinic is absent from them, it is absent from the layer the engines read. If it is present and accurately described, it has a genuine path into the answer. None of these individually guarantees a citation; together they form the source pool the engines sample from.

One detail rewards a second look. The sources with the broadest reach across engines are the neutral, structured ones: the CMTO register, Google's map data, and the massage-specific directories. The promotional, story-heavy asset, your own clinic site, is the one thing you fully control and the one thing the engines are least likely to cite unaided. That inversion is the core frustration of this vertical, and it is why the work cannot stop at redesigning the website. We dig into the map-data mechanism on its own in our look at the local signals that decide AI near-me answers, and into why third-party pages so often win in why directories dominate AI local search.

Each engine reads a different slice of the web, so this is four jobs

The engines do not share a source list, so "getting recommended by AI" is really four separate jobs wearing similar chat boxes. ChatGPT leans heavily on Google and Bing map data plus review platforms like Yelp, so a complete Google Business Profile and consistent NAP data are its levers. Perplexity tends to spread across several sources per answer, mixing a directory list with a couple of clinic sites and niche aggregators. Gemini often routes through Google's own grounding pipeline, which favours structured, schema-marked pages. Claude leans editorial, preferring human-curated "best of" shortlists it can relay.

Set those four retrieval habits next to each other and the conclusion is hard to dodge for a clinic owner. A clinic that optimised perfectly for the sources ChatGPT trusts would have done little for the engine that went straight to a curated listicle it never appears on. No single leaderboard spans ChatGPT, Claude, Gemini, and Perplexity, so there is no one ranking for an RMT to win. There are four retrieval systems, each grounding in a different corner of the web, and we unpack that fragmentation across every vertical we have studied in the cross-engine consensus gap. For an RMT, the practical response is not to chase one engine. It is to be accurately present in the neutral sources all four tend to reach for, then earn the platform-specific listings each one favours.

Why CMTO advertising rules quietly shape what AI can cite

This is the part that separates massage therapy from most local verticals, and the reason the counterintuitive claim at the top holds. CMTO advertising requirements are strict and specific: your advertising must be factual and verifiable, must clearly identify the RMT responsible for the practice, cannot use testimonials that exaggerate results, and cannot claim a specialty, title, or superiority you do not hold. These rules bind you online exactly as they do offline. Your website copy, your Google Business Profile description, and even your replies to reviews are all advertising under the College's definition.

For a marketing team, that reads like a constraint. For an AI engine, it reads like a gift. Generative models preferentially ground claims that are concrete, specific, and attributable, and they deprioritise vague promotional puffery precisely because it is hard to verify and easy to hallucinate around. The GEO research bears this out: the tactics that lifted AI visibility most were adding real citations, real quotations, and real statistics, not adding adjectives. CMTO has been forcing Ontario RMTs to write exactly that kind of factual, verifiable copy for years. Compliance and citability point the same direction. The rules you might have resented are the rules that make your material machine-preferred.

There is a trap Matt sees clinics fall into. A directory calls the practice the "top-rated" RMT in town, and the clinic, thrilled, puts "voted best massage therapist" on its homepage. That sentence, on a page the RMT controls, is the exact superlative the College forbids, and it can make the page less citable, since engines discount unverifiable promotional claims. The directory can say it. You cannot repeat it. Getting that boundary right is where regulated-vertical GEO earns its keep.

What a single-location Ontario clinic can actually do

A single-location practice wins AI visibility by earning a clean, accurate, consistent presence in the neutral sources the engines already read, then structuring its own site so a model can extract and attribute it. You are not trying to out-rank a national algorithm. You are trying to be present and legible in a short list of sources, a narrower and far more achievable job that maps directly onto the stack above.

For a regulated massage practice, two of Formative Digital's 12 Vectors do most of the heavy lifting. Vector 5, Cite, earns and corrects your placement in the third-party sources each engine trusts, which for RMTs means the CMTO register, Google's map data, RMTAO, and the massage directories. Vector 10, Localize, makes your local entity unambiguous to every retrieval system at once through consistent signals and local schema. We run these through the Formative Forces, our orchestrated multi-agent system, so one practice is worked across all four source layers in parallel rather than one engine at a time. The mechanics in a different, non-regulated field sit in our Mattress Miracle case study, where structured, citable content moved a Brantford retailer from roughly 1,000 to more than 82,400 monthly organic visits (SEMrush, April 2026). Massage therapy is a harder, health-regulated vertical, and outcomes depend on your competition, your city, and your existing presence, so treat that number as direction, not a promise. If you would rather have this done for you than done to a checklist, that is what our Formative Digital GEO services exist to handle.

How to measure your AI visibility across four engines

You measure it per engine, not as a single score, and most practices get this wrong by chasing one number. AI visibility is not one figure; it is four, one per engine, and they will disagree with each other. The method is to run the real client query, "best registered massage therapists in {your city}, Ontario," through ChatGPT, Claude, Gemini, and Perplexity on a schedule, record which practices each engine names and in what order, and track your share of those mentions over time. A Google ranking report tells you almost nothing here, because the engines barely read Google's ranked links; they read the map data and the directories instead.

Two refinements matter. First, monitor the register entries, map data, and directory listings feeding each engine, not only the chat output, since that source layer moves before the answers do. If you newly appear in a directory an engine favours, expect movement in that engine before anywhere else; the listing changes first and the citation follows. Second, sample each engine more than once, because AI answers carry real run-to-run variance, so a single check on a single day is weak. Score each of the four engines on its own, resample on a schedule, and the resulting read on a clinic's AI visibility is one you can actually trust. We lay out the full method in our guide to diagnosing your AI visibility across engines, and the trust mechanics behind it in how third-party citations build AI trust.

None of this is a trick, and nobody can promise a given engine will name a given clinic on a given day. The engines shift and the directories reshuffle. What the evidence supports is direction: the citation goes to retrievable, attributable, compliant sources, so an RMT who corrects the register, completes the Google profile, earns a clean place in the massage directories, and keeps every word factual competes for the AI answer far better than one still polishing a website the engines will not open.