We explained the methodology behind the AI Recommendation Score in an earlier post: five weighted categories, combined into a single 0 to 100 number and letter grade. That post covered the framework in the abstract. This one walks through an actual score, using St Mellion Estate's real test run, so a Grade C means something concrete rather than an idea of a grade.
St Mellion scored 68 out of 100, a Grade C. Here is exactly how that number was built, and what each part of it says about where the resort stands.
AI Recommendation Presence: 72, weighted at 35 percent
This is the category that decides most of the final score, because it measures the thing that actually matters commercially: does AI name the business when a real customer asks. St Mellion scored well here, mentioned on every one of eight golf focused prompts across Gemini and ChatGPT, ranked first on several. The score is not higher than 72 because of one clear structural gap: Perplexity, tested directly, did not mention St Mellion at all on the prompts re-verified live, including one where the other two platforms both called it the outright standout. A 35 percent weighted category can be pulled down meaningfully by a single platform behaving completely differently to the others, and that is exactly what happened here.
Website AI Readiness: 55, weighted at 20 percent
This category checks whether a model can actually parse what the business offers from its own website: crawlability, schema markup, plain text content, and NAP consistency. St Mellion's site is clean and crawlable, with no blocking of AI bots and real HTML content rather than JavaScript rendered pages. But there was no llms.txt file, no confirmed schema markup, and no FAQ content answering the exact style of question the panel test was asking. Three concrete, fixable gaps, none of them requiring a rebuild, all of them dragging a 20 percent weighted category down to the middle of the range.
Brand Authority: 76, weighted at 20 percent
The strongest category, and the one where the resort's genuine standing shows up clearly: Golf World Top 100 World Resorts recognition, regular golf trade press coverage, and an active social media presence. The one factor holding it below 80 is mixed review sentiment on the accommodation side specifically, a split both Gemini and ChatGPT picked up independently and reflected in their answers. Strong external reputation signals carry real weight in this category, and St Mellion has genuine ones to draw on.
Content Quality: 65, weighted at 15 percent
Concrete, citable facts exist across the site: course design credentials, room types, membership structure. The gap is that pricing specifically is inconsistent across third party sources, meaning no single canonical green fee figure exists for a model to cite with confidence, and much of the genuinely specific detail sits on subpages rather than the homepage, which leans more on marketing language than fact.
Technical Health: 70, weighted at 10 percent
The smallest weighted category, and correctly so, since it is the most generic of the five and the least directly tied to whether AI actually recommends the business. Clean page loads, a mobile viewport tag, and no crawl errors kept this comfortably mid range, with a live PageSpeed check flagged as the one outstanding item to confirm the number with full confidence.
What a Grade C actually tells a business owner
The headline number on its own would tell St Mellion almost nothing useful. The category breakdown tells a very specific story: the golf product is winning wherever it's tested, brand reputation is genuinely strong, and the score is being held down by a cluster of readiness and content gaps that are individually small, quick fixes, plus one structural platform specific gap on Perplexity that needs investigating on its own terms. That is the actual value of a weighted, category level score over a single free platform check. It does not just say whether you're visible. It says exactly which lever to pull first.
Where to start
If you've only ever seen a single number from a free AI checker, that number is telling you far less than it feels like it is. A weighted breakdown across five categories, evidenced with real AI conversations, is what turns "you scored 68" into an actual list of what to fix and in what order.