Golf clubs may be increasingly exposed to how AI assistants describe and recommend venues, after a new study by AI Score found that course recommendations can vary sharply depending on whether a golfer uses Perplexity or ChatGPT.
AI Score, a company that audits how AI assistants describe and recommend businesses, has found that the answer a golfer gets depends heavily on which platform they use, with the same club appearing strongly on one assistant and disappearing entirely on another.
The company tested four UK multi-site golf operators – The Club Company, Crown Golf, Mytime Active and Get Golfing, using realistic buyer prompts such as “best golf club near Sevenoaks” and “best golf club in Cornwall,” run on both Perplexity and ChatGPT. The results showed sharp inconsistency between platforms for the same business.
St Mellion Estate in Cornwall, a Crown Golf resort with a Nicklaus-designed course and European Tour history, did not appear at all on Perplexity for “best golf club in Cornwall,” while ChatGPT ranked it fourth out of four with an accurate, positive description.
Nizels Golf & Country Club in Kent (The Club Company) appeared on both platforms but was ranked last each time, with both AIs naming other nearby clubs as the stronger choice.
David Mullins, founder of AI Score, said: “The findings point to a gap most operators have no visibility into. A business can rank well on Google and still be missing or misrepresented when a prospect asks an AI assistant the same question, and the outcome can change completely depending on which assistant they use. For multi-site operators, that inconsistency can mean material differences in enquiry volume between clubs in the same portfolio, for reasons that have nothing to do with the quality of the course. None of these operators knew any of this until we ran the test.”
“Perplexity is the weak link. Across every venue tested on a generic, unbranded buyer-intent prompt, real searches like “best golf club near Plymouth”, Perplexity either omitted the business entirely or buried it as an afterthought. Zero strong, prominent placements on a generic query across the whole panel. This held even for the strongest profile in the sample, St Mellion, that Gemini and ChatGPT both independently called the standout for that exact search. Perplexity left it out completely, not in the written answer, not even in the map listings.
“It’s not about club quality,” Mullins continues. “The same businesses that vanished on Perplexity reappeared, often glowingly and with accurate detail, the moment the prompt was branded or hyper-local, and on the other platforms tested. AI models can find and describe these businesses correctly. One platform in particular just isn’t surfacing them on the generic discovery query that mirrors how a real prospect actually searches, the exact query type that drives new enquiries.”
He added: ‘The gap isn’t random. It concentrates on the two query types that actually drive bookings: generic “best club in [area/region]” searches, and use-case searches like “venue for a work social” or “golf break recommendation.” Branded and hyper-local queries (“is [club] any good”, “golf near [specific landmark]”) perform well almost everywhere. The businesses are losing exactly the searches a new customer would type, not the ones existing customers would.
PLATFORM BY PLATFORM, ACROSS THE FULL PANEL
ChatGPT: Inconsistent rather than consistently weak. Present in most cases, but position swings from #1 to dead last depending on the venue and the exact phrasing. One venue in the sample was absent entirely.
Gemini: The strongest performer where tested. On a full 8 prompt panel for one venue, it placed top 3 on every single prompt, including #1 on three. Where the underlying reputation data (reviews, press, awards) is solid, Gemini surfaces it reliably.
Claude: Smaller sample, but the same pattern, strong and accurate when a venue is named directly, less reliable on the unbranded discovery query.