We covered St Mellion Estate's AI Recommendation Score in an earlier case study: a Cornwall golf resort that ranks first or near first on Gemini and ChatGPT for almost every golf related prompt, and is completely absent on Perplexity. That gap was about which AI platform a prospect happens to open. There is a second gap in the same test run that has nothing to do with platform at all, and it is arguably the more useful one for any golf resort that sells accommodation alongside the golf.
The prompt that flips the result
Across eight golf focused prompts, St Mellion's course is mentioned on every single one by Gemini, and named "best overall" or ranked in the top position by ChatGPT on several. Then the panel test asked one hotel specific question: best four star golf hotel in the South West. The result inverted completely. ChatGPT dropped St Mellion entirely, naming four other properties plus two honourable mentions instead. Gemini kept it in the list, but at fourth of five, behind two named competitors it had never mentioned on any golf prompt.
Same estate, same facts, same underlying business. The only thing that changed was the category of the question.
Why golf and accommodation get judged separately
A "best golf club" question pulls on golf specific signals: course design credentials, tournament and society hosting history, green fee information, golf press coverage. St Mellion has all of this in depth, including a Jack Nicklaus designed course and European Tour event history, and it shows in the results.
A "best golf hotel" question pulls on an entirely different signal set: star rating confirmation, room type and amenity detail, and critically, accommodation specific review sentiment from sites like Tripadvisor rather than golf specific coverage. Both ChatGPT and Gemini independently surfaced the same underlying pattern in their answers when asked directly about the resort: strong, confident praise for the golf, alongside a consistently softer, more qualified tone on the accommodation, described in one answer as "a little dated in places." The models are not guessing. They are reading the actual review sentiment split that exists across Tripadvisor and Golfshake, and reproducing it faithfully in a category specific answer.
What this means if your business sells more than one thing
Most golf clubs are not single product businesses. A resort sells rounds, memberships, corporate days and, often, accommodation. We have written separately about the gap between AI recommending a club for a round versus for membership, and the gap between a generic golf prompt and a corporate event prompt. This is the same underlying pattern again: AI treats each buying intent as its own question, with its own competitor set, and a business can win decisively on one while losing on another, with no visibility into the gap unless someone actually tests each category on purpose.
The commercial risk here is specific. A resort with a strong golf reputation can reasonably assume that strength carries over into hotel bookings. The test data says it does not automatically carry over at all. A prospect asking AI specifically about accommodation gets a different, harder question answered by a different set of signals, and if those signals are weak, the resort loses a hotel booking to a competitor it would comfortably beat on the golf question.
What closes the gap
- Hotel specific schema markup. Star rating, room count, amenities and pricing structured separately from the golf course schema, so a model can confirm accommodation facts independently of golf facts.
- A dedicated hotel landing page with concrete detail. Room types, capacity, star rating stated plainly, and what's included in a stay, rather than folding accommodation into general resort marketing copy.
- Active review management on accommodation specifically. Golf reviews and hotel reviews often live on different platforms and get very different levels of attention. The accommodation side needs its own monitoring, because AI models are reading it directly.
- An FAQ addressing the accommodation question directly. "Is St Mellion a four star hotel," "what's included in a golf break stay," answered in plain text gives a model something current and specific to cite, rather than leaning on older or mixed review sentiment by default.
Where to start
If your golf club or resort sells accommodation, corporate days, or membership alongside the core round, do not assume AI treats all of it as one business. Test each buying intent separately, the same way a real customer would ask about it, and expect the answer to be different for each one.