Kokomo Botanical Resort
LLM & GEO Analysis

1Executive summary

Right now, when a wellness traveler asks ChatGPT, Perplexity, Gemini, or Claude about Turks and Caicos, Kokomo is barely citable. Not because the brand lacks substance, there is plenty: a Caribbean Wellness Resort of the Year award, BioGeometry architecture, 20-year botanical gardens, Marriott Bonvoy Homes & Villas listing, a top-50 Caribbean restaurant, six cottage variants, a spa. The problem is mechanical: those facts are either locked inside meta tags AI models discard, or stated nowhere on the page in extractable form, or typed wrong in the structured data so AI classifies the property as a "Person" or a generic "Article" instead of a Resort.

Five sentences

1. Kokomo's positioning lives in metadata, not in body content. AI assistants read body content; the metadata gets stripped. The award narrative, the BioGeometry concept, the Marriott Bonvoy partnership, the 20-year gardens, the chemical-free wellness story, none of it is in homepage body text.
2. The schema graph mis-classifies the resort as a Person. Every page declares a Person + Organization root entity and types itself as Article. There is no LodgingBusiness, HotelRoom, Restaurant, DaySpa, EventVenue, FAQPage or AggregateRating anywhere. AI assistants prioritising hospitality types skip Articles.
3. The six cottage pages give an LLM no facts to cite. Zero square footage, zero rate ranges, zero "from $", zero ocean/garden view labels per cottage, zero structured amenity lists. "How much is a Luxury 2-King at Kokomo?" returns nothing because the page does not say.
4. The brand name is fragmented across signals, schema says "Kokomo Botanical Garden Resort," every other surface says "Kokomo Botanical Resort." Entity resolution in Knowledge Graph and Perplexity matches on schema first, so the canonical record is being split.
5. Three live spam posts in the sitemap are signaling distress to LLM crawlers. Casino + Microsoft Office warez content rendered under Kokomo's template, indexable, served at 200 OK. Some assistants now downrank or refuse to recommend domains carrying this pattern. See §10.

2What AI assistants see today P0

The clearest way to read this audit is to picture an LLM trying to answer the questions a wellness traveler actually types. Below is what is currently extractable from the live site for each question, based on the body content and JSON-LD parsed on 7 May 2026.

The pattern across all of the above is the same: the resort has the substance, but it is encoded in places AI does not read. The next sections walk through each cluster and what to fix.

3Five claims hidden from LLMs P0

Every key positioning claim Kokomo makes about itself appears in a <meta> tag, never in visible body content on the homepage. Modern AI assistants increasingly strip metadata before answering, they want body content, headings, and lists. So the resort gets very little credit for what the meta tags promise.

What to publish in body content (not meta tags)

1. Property fact panel, name, location with parish/island, opening year, parent ownership/affiliation, total cottages, total acreage, gardens age, distance to Grace Bay Beach, distance to PLS airport. Plain language, plain numbers, scannable list.
2. "What is BioGeometry?" explainer, one paragraph definition, one sentence on how it's applied at Kokomo, one citation link to a third-party source. Distinctive entities like this are exactly what LLMs love to summarise, but only if the page explains them.
3. Marriott Bonvoy trust block, small Bonvoy logo, a one-sentence explanation of what the partnership means for guests (loyalty points, customer service standards), link to the Bonvoy listing.
4. Awards section with provenance,"Caribbean Wellness Resort of the Year, [year], [awarding body]", and "TripAdvisor Travelers' Choice [year]." Link to the awarding body where possible.
5. Founder / owner page with photos, bio, philosophy, a quote LLMs can pull. Pure E-E-A-T move and currently the largest gap on the About Us page.

The About Us page, most striking single finding

The About Us page meta description reads "Award-Winning Wellness Resort in Turks & Caicos." Inside the page body, the words BioGeometry, Marriott, award, wellness as positioning, and founder/owner story are all under-used or absent. An About Us page is the single highest-leverage page for LLM authority signals. Right now it tells AI assistants almost nothing about who built Kokomo, why, when, and what makes it a wellness resort versus a beach resort.

4Cottages, six pages, no AI-extractable facts P0

Six cottage detail pages plus the parent /cottages/ hub. None of them give an AI assistant the facts a guest asks for.

Counts across all six cottage pages

Naming inconsistency: "Luxury" vs "Ultra-Lux"

Top-tier cottage URL slugs use luxury-... and the page titles say "Luxury 2BR Cottage." The main site navigation labels the same products "Ultra-Lux King and Two Twin Cottages" / "Ultra-Lux Two King Cottages." Pick one term, easiest path is to update the menu labels to match the URL/title ("Luxury"), since URL slug is the harder signal to change.

HotelRoom schema each cottage page should declare

{
  "@type": "HotelRoom",
  "@id": "https://kokomobotanicalresort.com/cottages/luxury-two-bedroom-cottage-2-king/#room",
  "name": "Luxury Two-Bedroom Cottage (Two Kings)",
  "description": "Adults-oriented two-bedroom cottage with two king beds, ...",
  "containedInPlace": { "@id": "https://kokomobotanicalresort.com/#resort" },
  "occupancy": { "@type":"QuantitativeValue", "maxValue": 4 },
  "bed": [
    { "@type":"BedDetails", "numberOfBeds":2, "typeOfBed":"King" }
  ],
  "amenityFeature": [
    { "@type":"LocationFeatureSpecification","name":"Wi-Fi","value":true },
    { "@type":"LocationFeatureSpecification","name":"Air Conditioning","value":true },
    { "@type":"LocationFeatureSpecification","name":"Kitchen","value":true },
    { "@type":"LocationFeatureSpecification","name":"Private Patio","value":true }
  ],
  "floorSize": { "@type":"QuantitativeValue", "value": 950, "unitCode":"FTK" },
  "image": [ "..." ],
  "url": "https://kokomobotanicalresort.com/cottages/luxury-two-bedroom-cottage-2-king/"
}

5Dining, spa, weddings, same problem P1

WE Market Cafe

Page mentions "menu" 7×, breakfast/lunch/dinner each 4×, but contains no operating hours, no signature dish, no chef name, no price range, no reservation policy, and no Restaurant / FoodEstablishment schema. The top-50 Caribbean restaurants accolade lives in one blog post, not on the restaurant page itself, not in schema. AI assistants asked "best restaurants in Grace Bay" have nothing to grab.

Lovina Spa

1 mention of "hours" without a timetable. 2 mentions of "menu" without a treatment list. 0 mentions of treatment names. No DaySpa schema. AI cannot answer "spa treatments at Kokomo," "spa hours at Kokomo," or "spa prices at Kokomo."

Wedding Venues

No capacity numbers, no venue layout types (beach ceremony / garden ceremony / indoor reception), no package starting prices, no sample timeline, no list of included services. AI assistants asked "wedding venues in Turks and Caicos under 80 guests" cannot match Kokomo because the page does not state a guest range.

Gatherings & Corporate Retreat

/gatherings/ and /kokomo-corporate-retreat/ are byte-identical except for canonical. To AI this looks like duplicate content under two URLs, fragments topical authority. Pick one URL or split into two genuinely distinct pages targeting "corporate retreat Turks and Caicos" and "weddings/gatherings Turks and Caicos" separately.

7Entity name fragmentation P1

For LLM entity resolution the canonical name of the property must be identical across every signal AI ingests. The site is sending mixed signals:

Schema names the entity Kokomo Botanical Garden Resort, every other surface drops "Garden." Knowledge Graph and Perplexity match on schema first; the canonical record is being split. Pick one canonical name (likely the domain-aligned "Kokomo Botanical Resort"), and use it identically in JSON-LD name, og:site_name, footer copyright, GBP, TripAdvisor, Marriott Bonvoy listing, and every social profile. The header kicker "Caribbean Family Cottages" sits above a wellness-resort H1 and adds a third positioning that AI has to reconcile, drop it or rewrite to a wellness-aligned line.

Cross-reference: the GBP-side damage from the same fragmentation, mobile-Maps truncation of the "- Caribbean Family Cottages" tail and classifier confusion with the family segment, is documented in GBP Analysis §2. The fix is two coordinated edits: this file's schema rewrite plus that file's GBP profile rename.

9Image alt text & visual AI search P2

Across the 33 audited pages, ~93% of <img> tags ship with empty (alt="") or missing alt attributes. Per-page counts in §12.

Why this matters specifically for LLMs:

• ChatGPT image search, Perplexity image carousels, Gemini image grounding rank heavily on alt text and surrounding caption text.
• Google Lens and Apple Visual Look Up use alt as the primary signal for "where can I stay that looks like this."
• Pinterest's automatic save uses alt as the default pin description, a major top-of-funnel loss for a visually-driven wellness resort.

Format mandate: "[Subject] at Kokomo Botanical Resort, Providenciales,[short context]." Examples: "Two-king cottage interior with botanical garden view at Kokomo Botanical Resort, Providenciales, soft morning light"; "Lovina Spa massage room at Kokomo Botanical Resort, Providenciales, stone wall, plants, candle lighting." Decorative images keep alt="" only after a deliberate audit.

11Sitemap, metadata & canonical hygiene

Compressed coverage of the supporting hygiene issues that matter for LLM crawl quality. None of these are headline findings, but each one is a small drag on AI-side trust and indexation.

Sitemap inventory

Pages that need title or description rewrites

Robots / canonical issues

• /thank-you/ is index, follow. Should be noindex, follow.
• /wellness-retreat/ and /kokomo-corporate-retreat/ are indexable with canonical pointing elsewhere. Replace with 301.
• Site uses en-CA in HTML and schema, while og:locale says en_US. Pick one, for TCI tourism, en-US is the more common AI-assistant default.
• /blizzard-snow-storm-policy/ on a Caribbean resort, repurpose or 410.

12Per-page findings table

Word-count is from rendered body text. "Top issues" lists the LLM-relevant per-page items.

13Roadmap

Sequenced for AI-visibility impact-per-hour. Built to be implemented inside the planned site rebuild rather than retrofitted onto pages that will be replaced anyway.

P0, Stop bleeding This week

1. Remove the 3 spam posts (410 + Search Console removal). Audit for hidden ones. Full security cleanup. (§10)
2. Replace the misconfigured root schema with a single Resort + LodgingBusiness entity, address, phone, email, awards, aggregateRating, sameAs. (§6)
3. Convert the existing visible FAQ blocks (homepage, /wellness/) into FAQPage schema. (§6)
4. Add og:image + Twitter card to every page. (§8)
5. Publish the five hidden claims (BioGeometry, Marriott Bonvoy, awards with provenance, gardens age, chemical-free) in body content on the homepage and About Us page. (§3)
6. Pick one canonical brand name and unify across schema, og:site_name, GBP, Bonvoy, social. (§7)

P1, Bake into the rebuild Compound the win

1. Add HotelRoom schema to all 6 cottage pages with bed configs, occupancy, view type, amenityFeature, floorSize, priceRange. Publish those facts in body content too. (§4)
2. Add Restaurant schema to /we-market-cafe/ with hours, servesCuisine, priceRange, award, menu URL. Publish hours and signature dishes in body. (§5)
3. Add DaySpa schema to /lovina-spa/ with hours, treatments, priceRange. Publish treatment list in body. (§5)
4. Add EventVenue schema to /wedding-venues/ and /gatherings/ with maximumAttendeeCapacity, layouts, priceRange. (§5)
5. Resolve cottage label inconsistency (Luxury vs Ultra-Lux). (§4)
6. Fix Casitas subdomain: title, description, canonical, og:image, robots; reduce to 1 H1; link properly from main site. (§2)
7. Bring all six cottages into Marriott Bonvoy Homes & Villas; link each cottage CTA to its Bonvoy listing. (§2)
8. Audit alt text site-wide. (§9)
9. 301-redirect duplicate URLs; noindex thank-you / terms / privacy. (§11)

P2, Build authority 90 days

1. Founder / owner page with photos, bio, philosophy, citation-ready quote. (§3)
2. Press & media coverage page consolidating all awards with publisher and link, each schema'd as CreativeWork.
3. Distance panel (PLS airport / Grace Bay / Provo Golf Club) with map embed and Place schema.
4. Treatment menu published on Lovina Spa page with name, duration, price band, Service schema per treatment.
5. Wedding capacity calculator + sample timelines + price-band packages.
6. Internal linking pass: every blog post links to 2–3 most-related cottage / spa / dining / weddings pages.
7. Pillar pages, wellness travel, Turks & Caicos guide, corporate retreat, each interlinking related posts.
8. Languages-spoken declaration in schema and on contact page.
9. Quarterly schema-validation in Search Console.

P3, Backlog When the basics are paid

1. Localised landing pages (ES / FR / IT) for European markets, with hreflang.
2. Image alt-text generation pipeline retrofitted to historical media library.
3. VideoObject schema for property-tour and Lovina Spa videos.
4. Podcast appearance inventory with timestamps and pull-quotes.

14Methodology

Crawled

• https://kokomobotanicalresort.com/sitemap_index.xml and the five child sitemaps (post, page, promotion, archive, category).
• 33 core URLs from the page-sitemap as the audit sample, plus the homepage.
• Casitas subdomain at https://casitas.kokomobotanicalresort.com/.
• The three spam posts to verify HTTP status, robots tag and rendered language.

Per-page checks

• HTTP status, response size, redirect chain, canonical, robots, hreflang.
• Title and meta description length and content.
• Open Graph and Twitter Card tags.
• JSON-LD blocks, type, key fields, validation.
• H1, H2, H3 hierarchy and duplication.
• Body word count, entity keyword presence, NAP signals.
• Image count, missing alt count, lazy-loading attribute.
• Internal vs external link counts.
• FAQ presence in HTML (Q?H question pattern).
• Booking-engine link presence and target.

Out of scope

• Mobile UX, design quality, conversion-rate audit, covered in Kokomo Website Analysis 2026.
• Core Web Vitals (LCP/INP/CLS), separate live-test discipline.
• Backlink profile / toxic-link cleanup, recommended Q3 2026 after schema is fixed.
• Google Business Profile audit, recommended in parallel.
• Localized variants (ES / FR / IT).

Tooling

Live-site fetch, sitemap parsing, JSON-LD extraction, body-text normalisation and entity-keyword counts performed against snapshots taken on 7 May 2026. Findings verified by re-fetching anomalous pages a second time. Where this report quotes a count (e.g. "BioGeometry 0× in body"), the count is from the rendered, script-stripped, style-stripped DOM text, what an LLM would actually parse.