DUBLIN – 24 September, 2026 – Patrick McKenna, the Dublin-based AI search specialist and founder of Canvas Marketing Solutions, today released results from his work with a Perth cosmetic clinic. Google Search Console data shows the clinic’s AI search visibility grew more than sevenfold in three months. McKenna also published an analysis of a ChatGPT search in which the assistant recommended the clinic over 12 competitors, along with The GEO Playbook, his free 23-tactic guide to earning visibility in AI answers.
From 42 to 324 AI appearances a day in 90 days
Between 21 June and 20 September 2026, the clinic’s website appeared 17,483 times in Google’s generative AI features, including AI Overviews and AI Mode.
Daily AI impressions averaged 48 in late June, 108 in July, 218 in August and 345 in September. The first week averaged 42 a day and the final week 324, an increase of more than sevenfold. The highest single day, 9 September, recorded 422 impressions.
Conventional search grew alongside AI visibility. The site earned 1,471 clicks from 55,286 Google search impressions over the period. Average daily clicks tripled, from about 8 in the first fortnight to 26 in the last. Average ranking position improved from 17.2 to 9.3.
“Most businesses still judge their website on clicks alone,” said Patrick McKenna. “But more and more, the first moment a customer meets a business happens inside an AI answer. For this clinic, that moment went from about 40 times a day to more than 300 in a single quarter. That’s the metric I now watch most closely for every client.”
Inside a ChatGPT recommendation
McKenna analysed an export of a ChatGPT search asking for the best place in Perth for the clinic’s treatment. The assistant’s reasoning text was hidden in the export, so his analysis is based on the searches it ran, the pages it opened and the sources it cited.
The assistant first rewrote the user’s question into its own search query, which returned 18 results from 13 Perth clinics. In its very next round of research, it narrowed the field to two clinics and opened the full websites of only those two. Every other clinic was judged on a one-line search snippet.
“Thirteen clinics made the first cut, and eleven were gone before the AI read a single full page,” said McKenna. “That’s the part most marketers haven’t grasped yet. You don’t get a fair hearing at the end of the process. You win or lose in the first few seconds, on how clearly your snippet answers the question.”
McKenna identified three factors that decided the recommendation.
1. Matching the words customers actually use. The user typed a non-standard spelling of the treatment rather than the industry term. McKenna had built the clinic’s online identity around exactly that wording, so its site was the natural match. Search Console confirms the pattern: the site ranked at an average position of 2.46 for one of those customer-spelled variations.
“I build every site around the words customers type, not the words the industry prefers,” said McKenna. “People don’t search in textbook terminology. If your website only speaks the industry’s language, AI assistants will pass you over for someone who speaks the customer’s.”
2. A named expert, not an anonymous business. The assistant said it would compare practitioner credentials. Of the snippets it saw, the clinic’s was the only one to name its practitioner and describe her experience. Most competitors described only the treatment.
“AI assistants are looking for a person to trust, not just a service,” said McKenna. “Put a real, named expert with a verifiable history front and centre, and you immediately stand out from competitors whose websites could belong to anyone.”
3. Carrying a decades-long reputation to a new business. When the assistant searched the practitioner’s name, it found her previous Perth clinic, which she had since sold, and its 4.9-star rating from 90 Google reviews. Many of those reviews praised her personally.
“The practitioner had spent more than 30 years building her reputation before selling her previous clinic,” said McKenna. “The business changed hands, but the reviews were about her: her skill, her care, her results. A brand-new website normally starts from zero trust. This one didn’t, because I presented her consistently as the same person, with the same name, history and credentials. When the AI followed her name, it found one consistent expert with decades of track record behind her.”
He added: “Businesses get sold, brands change and websites get rebuilt. A person’s reputation is the one asset that survives all of that, but only if search engines and AI systems can connect that person to what they’re doing now. Making that connection is exactly the work I do.”
Press releases feeding AI answers
The ChatGPT breakdown also showed the assistant surfacing a press release during its background research. This is part of a process known as query fan-out, in which an assistant splits one question into several related searches.
“This is what I’ve been seeing when I test customer questions in AI assistants for my clients,” said McKenna. “Details from press releases I distribute are being picked up in those fan-out searches. An AI answer is assembled from evidence across the web, not just your own site. Genuine news in the right places becomes part of what the AI knows about you.”
When verification fails, AI relies on what businesses publish
McKenna’s analysis also found that many of the assistant’s attempts to check clinics independently came back empty. Its review searches returned irrelevant results, including unrelated organisations with similar names and scientific articles.
“That’s the finding every business owner should pay attention to,” said McKenna. “When an AI can’t find independent evidence, it leans heavily on what a business says about itself. That’s an opportunity, but it’s also a responsibility. Your information has to be accurate, current and honest, because sometimes it’s the main thing the AI trusts.”
He was clear about the limits of one search. “This was one search on one day. AI answers change from person to person and hour to hour, so I’d never promise anyone a guaranteed AI recommendation. What the breakdown shows is which signals the AI actually weighed, and those are signals any honest business can build.”
A website written line by line with AI
McKenna built the clinic’s website in code, line by line, using AI-assisted development instead of a template or page builder.
“I wrote every line of that site with AI as my development partner, so nothing on it is there by accident,” said McKenna. “There are no bloated plugins and no template code getting in the way. AI systems get clean structure and clear answers, and people get plain-language information when they need it.”
The data suggests AI systems reward that answer-first approach. Google cited 62 different pages from the site in AI answers. The most-cited was a guide to healing time, with 4,022 AI impressions, followed by a day-by-day aftercare guide with 3,486, the homepage with 2,060, and a guide to how long results last with 1,430. One treatment comparison page drew 77% of its total search visibility from AI answers: 486 of its 635 impressions.
“The pages AI cites most are the ones that answer the question a nervous customer asks at 11pm after booking,” said McKenna. “How long will this take to heal? What happens on day three? Is it safe? Answer those questions honestly and in detail, and AI starts treating you as a source rather than an advert.”
The site’s AI visibility reached well beyond Western Australia. Australia accounted for 36% of AI impressions, and the site appeared in AI answers across 151 countries. Mobile searches made up 76% of AI appearances.
“A clinic in Perth being cited in AI answers in 151 countries tells you how differently this new kind of search works,” said McKenna. “Authority on a topic travels further than a local business listing ever could.”
The GEO Playbook: practical, ethical, testable
McKenna’s The GEO Playbook: How to Earn Visibility in AI Answers sets out 23 tactics across four areas: foundation, content, amplification and distribution, and measurement. It covers mapping query fan-out, building independent third-party proof, defining a person or brand as a clear entity, and tracking AI visibility with a fixed panel of test prompts.
The guide states plainly that no tactic guarantees an AI citation, and warns businesses against providers who promise “force indexing,” fake traffic or manufactured reviews.
“There’s a lot of noise in AI search right now and a lot of promises nobody can keep,” said McKenna. “These results came from unglamorous work: answering real customer questions, building a technically clean site, carrying a genuine reputation into a new business, and getting real news in front of real publications. I’ve written it all down so any business can see exactly what I do, and hold any provider, including me, to that standard.”
The GEO Playbook is available free at https://www.canvasmarketingsolutions.com/geo-playbook
About Patrick McKenna
Patrick McKenna is a Dublin-based AI search and digital marketing specialist and the founder of Canvas Marketing Solutions. He spent years working on Google Ads contracts across the UK and Ireland before becoming an independent marketing consultant in 2019. He now specialises in generative engine optimisation, SEO and AI-assisted web development. He has also developed AI-powered marketing tools, including Janice AI for automated blog content and BacklinkProspector.AI for backlink outreach. McKenna works with businesses across Ireland, the UK and Australia, with particular expertise in the beauty and aesthetics sector.
About Canvas Marketing Solutions
Canvas Marketing Solutions is a Dublin-based full-stack digital marketing agency founded by Patrick McKenna. It offers SEO, generative engine optimisation, Google Ads and Meta Ads management, AI-assisted web development and PR distribution.
Media ContactCompany Name: Canvas Marketing Solutions LTDEmail: Send EmailPhone: 0896040336Address:82 The Dunes City: PortmarnockState: DublinCountry: IrelandWebsite: https://www.canvasmarketingsolutions.com/
