St Mellion Estate is a well known golf resort in Saltash, Cornwall, part of the Crown Golf group, with European Tour event history and a Jack Nicklaus designed course. It is exactly the kind of established, well reviewed business that would assume it shows up wherever people are looking for it. We ran a full panel test to check, using real buyer style prompts across Perplexity, Gemini, ChatGPT and Claude.
The result was not "strong" or "weak." It was platform specific, and that turned out to be the actual story.
The pattern
| Prompt | Perplexity | Gemini | ChatGPT |
|---|---|---|---|
| Best golf club near Plymouth | Absent, four competitors named instead | Ranked #1 of 4, "premier golfing destination" | Named "best overall," described as the standout pick |
| Best golf club in Cornwall | Absent, five competitors named, not listed even as an alternative | Ranked #3 of 5 | Ranked #4 of 4, mentioned accurately |
| Golf club for a society day in Cornwall | Not re-verified this pass | Ranked #1 of 4, listed first with specific package detail | Ranked #4 of 4, mentioned accurately |
| Best 4 star golf hotel in the South West | Ranked #4 of 5 | Absent, top four plus two honourable mentions named instead | Mentioned, weaker framing than location based prompts |
Read the two Plymouth rows again. Gemini and ChatGPT both independently call St Mellion the standout choice for the exact same question. Perplexity does not mention it at all, and instead names Boringdon Park Golf Club as "best all around," despite St Mellion being the area's most prominent course by profile.
Why this happens
Each AI assistant builds its answer from a different mix of sources. Some lean heavily on a narrower set of indexed review and map sources. Others pull more from general web crawling and structured data on the business's own site. A business can be strongly indexed in the sources one model favours and effectively invisible in the sources another model favours, with no single obvious cause and no single fix that covers every platform at once.
That is exactly why testing across all four major platforms matters more than checking one and assuming the result generalizes. A business that only checks ChatGPT, for example, would have walked away from this test believing everything was fine.
What the score also found
Beyond the platform split, the panel test surfaced two other concrete, fixable gaps. Green fee pricing was inconsistent across third party sources, meaning no single canonical figure exists for AI models to pull from confidently. And no structured FAQ content exists on the site answering the exact style of question the AI panel was being asked, such as whether the resort suits beginners or what is included in a golf break, which is precisely the kind of plain language content AI models lift directly into an answer when it exists.
The takeaway
A business can be doing well by every traditional measure, strong reviews, real accolades, established brand, and still be structurally invisible on one major AI platform while dominant on another. The only way to know is to test the actual prompts a real customer would use, across every platform that customer might open, not just the one your team happens to check.