Hawaii AI Search
Visibility Index
— 2026

1. Why an Open Methodology Framework?

AI search is no longer hypothetical. ChatGPT has crossed 200 million weekly users. Google AI Overviews now appear on a meaningful share of search result pages. Perplexity, Claude, Gemini, and Copilot each route queries to recommended businesses. Hawaii visitors — particularly the 9.6 million annual non-resident travelers — increasingly use these platforms to plan their trips, find providers, and select vendors before they ever land.

The visibility decisions made inside those platforms are consequential and opaque. A Hawaii hotel that ranks first on Google may not be cited at all by ChatGPT for "best Maui resort." A Honolulu specialist with strong reviews may be invisible in Perplexity's response to "where should I go for [condition] in Hawaii." Until now, no public framework has standardized how to measure this — every agency, journalist, and analyst with a take on AI search visibility is using different queries, different engines, and different scoring rules.

The Hawaii AI Search Visibility Index closes that gap by publishing a single open methodology and applying it as an evolving dataset. Volume 1 (May 2026) is the first applied measurement: 50 queries against Anthropic Claude, with results cross-validated against verifiable Hawaii business sources. The methodology is the long-lived artifact; volumes are time-stamped applications of it.

The methodology is published under Creative Commons BY 4.0. Refinements happen openly and prospectively; the rules in effect at the time a dataset is produced are the rules that govern that dataset's interpretation. Anyone can apply the framework — Nekko Digital does in Volume 1, other agencies and researchers are welcome to extend it.

2. Methodology

The framework is built on a fixed query protocol designed to produce comparable, reproducible measurements across AI engines and across time.

Query universe

Each category contains 50–200 representative high-intent queries — phrased the way real Hawaii visitors and residents actually search. Queries cover destination-stage research ("best places to stay on Maui"), specialty + location patterns ("dermatologist Honolulu"), comparison patterns ("Wailea vs Kapalua for a wedding"), and direct-booking patterns ("snorkel tour Lanai with kids"). Volume 1 applies a 50-query subset; the full universe activates in subsequent volumes.

AI engines tested (full framework)

• ChatGPT — latest available GPT model with web search enabled. Pending in Volume 1; activates Volume 2.
• Perplexity AI — default mode with web search. Pending in Volume 1; activates Volume 2.
• Google Gemini / AI Overviews — both standalone Gemini queries and the AI Overviews surface inside Google Search. Pending in Volume 1; activates Volume 2.
• Anthropic Claude — web-enabled mode where available, otherwise tool-augmented runs. Active in Volume 1 — training-data baseline (no web augmentation).
• Microsoft Copilot — supplementary measurement for cross-platform comparison. Pending in Volume 1; activates Volume 2.

Run protocol

The framework specifies five runs per query per AI engine to account for non-determinism. Runs are spaced across multiple days to capture temporal variance. Each run captures: full response text, business names cited, citation position, source URLs cited, model version, and timestamp. Sessions are isolated — no carryover context between queries. Volume 1 uses single-run sampling per query as a baseline; Volume 2 commits to the full five-run protocol.

Data cleansing

Captured responses are cleansed to remove ad-style content, hallucinated business names (verified against Hawaii Business Express and DCCA records), and duplicate citations within the same response. Disambiguation rules handle similarly-named businesses and chain locations. Volume 1 cross-validated cited businesses against TripAdvisor verified property pages, the Honolulu Board of REALTORS market data, and hospital / health-system websites.

Methodological constraints

The framework deliberately excludes pay-for-inclusion mechanics, curated "best of" submissions, and any sample weighting that favors specific businesses. The query universe is fixed at the time of any data run — researchers cannot adjust queries mid-run to favor specific outcomes. New categories or query expansions are added with a clearly marked first-application flag so readers of any dataset know what is new versus continuing data.

3. What Categories Does the Framework Cover?

The framework documents methodology for five core categories chosen for their relevance to Hawaii's economy and the high volume of AI-routed visitor and resident queries each receives.

Core five categories

• Hospitality. Hotels, restaurants, and tour operators — the largest and densest citation-bearing category. See the Hawaii Hotel Marketing Guide for context on this segment.
• Real estate. Brokerages and individual Realtors per island, with submarket-level depth in Honolulu, Wailea, Kona, and Princeville. See the Hawaii Real Estate Marketing Guide.
• Healthcare. Hospital systems, specialty practices by specialty + location, and primary care visibility. See the Hawaii Healthcare Marketing Guide.
• Weddings. Venues, planners, photographers, florists, officiants. See the Hawaii Wedding Marketing Guide.
• Home services. Contractors, landscaping, pest control, cleaning. Coverage is thin in Volume 1 — see Volume 1 limitations.

Extending the framework to additional categories

The methodology is extensible. Applying it to professional services (legal, financial, accounting), retail and boutique, education, automotive, or fitness uses the same five-step protocol — fixed query universe, multi-run sampling, cross-engine measurement, scoring rubric, dataset publication — with category-specific query lists. Any party extending the framework to a new category is expected to publish the category-specific query universe alongside their dataset, so the work remains replicable.

4. Scoring Rubric

Each business measured under the framework receives scores across five dimensions. Composite scores combine the dimensions with documented weighting; the dimensional scores are also reported independently so readers can apply their own weightings.

• Citation rate. Percentage of relevant query runs across all engines where the business is mentioned. The headline metric — the share of times the AI brings up your name when asked something you should be the answer to.
• Citation position. When cited, how prominently — first mention, second, mid-list, or trailing. Lead citations carry more weight than mid-list mentions in user attention research.
• Citation context. Whether the AI engine frames the citation positively, neutrally, or negatively. Negative framing counts as visibility but not as a positive signal.
• Source attribution. Whether the AI engine cites the business's own URL as a source vs only mentioning the name. Most actionable visibility outcome.
• Cross-platform consistency. Whether the business is cited across multiple AI engines or only one. Limited applicability in Volume 1 (single-engine baseline); activates fully in Volume 2.

5. Volume 1 Dataset — Anthropic Claude Baseline (May 2026)

Volume 1 is a single-engine pilot dataset: a 50-query universe applied to Anthropic Claude (Sonnet, training-data baseline, no web augmentation), measured in May 2026. Results below are the businesses Claude cited when asked the queries in each category, cross-validated for existence and current operating status against publicly verifiable Hawaii sources. Volume 2 expands to multi-engine, multi-run coverage.

5.1 Volume 1 scope and disclosures

A note on what Volume 1 is, and what it is not. Volume 1 measures Claude's training-data familiarity with Hawaii businesses — which businesses surface when Claude is asked typical Hawaii intent queries. This is a real and useful signal: a business that surfaces strongly in a major AI engine's training corpus likely has accumulated meaningful third-party authority (press, reviews, citations, structured data). But it is one signal among many. A business absent from Volume 1 is not necessarily invisible in AI search broadly — it may surface strongly when the same query is run with web augmentation, against a different engine, or in a different time period. Volume 2 will measure those dimensions.

What is included in Volume 1: Citations of real, verifiable Hawaii businesses that Claude produced organically against the open query universe. Each cited business was validated against a publicly checkable source (TripAdvisor verified property page, Honolulu Board of REALTORS data, hospital / health-system website, or equivalent).

What is excluded from Volume 1: Businesses Nekko Digital was actively working with at time of publication (to maintain conflict-of-interest hygiene); businesses Claude cited that could not be verified as currently operating; hallucinated business names; advertising-style content; duplicate citations within a single response.

How to read the tables below: Rank reflects citation frequency across the relevant query subset, with citation position used as a tiebreaker (lead citations weighted more heavily than trailing mentions). Notes columns reflect Claude's contextual framing — the language Claude used when citing the business. Cross-validation columns reference where each business's existence and current status can be verified.

5.2 Hospitality — Hotels (10 queries)

Query subset: "Best hotels in Waikiki" · "Top luxury resorts in Maui" · "Best hotels for families in Hawaii" · "Where to stay on the Big Island" · "Best hotels for honeymoon in Hawaii" · "Most luxurious resort in Hawaii" · "Best hotels in Kauai" · "Boutique hotels in Hawaii" · "Best beach resorts in Hawaii" · "Top hotels for couples in Maui."

Patterns observed. Brand-anchored luxury (Four Seasons, Halekulani, Aulani, Royal Hawaiian) dominates citations. The pattern is consistent with how AI training data weights brand authority signals — recurring press coverage, awards databases, TripAdvisor depth, and structured data on flagship properties accumulate over time. Independent boutique properties without sustained third-party press are systematically underrepresented. This is also where a hotel's marketing strategy matters most: AI search visibility tracks accumulated authority, not current rate-tier.

5.3 Hospitality — Restaurants (10 queries)

Query subset: "Best restaurants in Honolulu" · "Best restaurants in Maui" · "Where to eat in Waikiki" · "Best Hawaiian food restaurants" · "Best plate lunch in Hawaii" · "Best fine dining in Hawaii" · "Best sushi in Hawaii" · "Best seafood restaurants in Hawaii" · "Top chef restaurants Honolulu" · "Best brunch in Honolulu."

Patterns observed. Hawaii Regional Cuisine chefs (Roy Yamaguchi, Peter Merriman, Alan Wong, Sam Choy) earn restaurant citations that propagate across queries through their multi-location brands. The plate-lunch and traditional Hawaiian food category is dominated by long-tenured local restaurants (Helena's, Highway Inn, Rainbow Drive-In) — citation rate appears tied to longevity and James Beard / press recognition, not current operational metrics. Mama's Fish House on Maui is uniquely dominant — it surfaces as a top citation across an unusually broad query range, suggesting accumulated press authority that extends beyond its Maui geographic context. Detailed restaurant marketing dynamics in Hawaii are covered in our restaurant marketing page. Note: training-data freshness limits — verify current operating status before citing in published downstream work.

5.4 Hospitality — Tour Operators (8 queries)

Query subset: "Best snorkel tours in Hawaii" · "Best whale watching tours Maui" · "Best helicopter tour Hawaii" · "Things to do on the Big Island" · "Best activities in Kauai" · "Sunset cruise Waikiki" · "Best boat tour Hawaii" · "Hawaii zipline tours."

Patterns observed. Activity-niche dominance — single operators control disproportionate citation share within their niche (Pacific Whale Foundation in whale watching, Body Glove in Kona snorkel, Atlantis in submarine, Blue Hawaiian in helicopter). This concentration is a function of accumulated review depth and press authority. Cross-validation against TripAdvisor's Hawaii top-10 attractions confirms the destinations driving these activities (Diamond Head, Pearl Harbor, Volcanoes National Park, Haleakala) but the operator-level citations come from training data rather than the destination listings themselves. New operators face a meaningful authority moat. Strategic context for this category is in our tour operator marketing page.

5.5 Real Estate (7 queries)

Query subset: "Top real estate brokerages in Hawaii" · "Best Realtor Honolulu" · "Luxury real estate Hawaii" · "Hawaii real estate market 2026" · "Best brokerages on Maui" · "Real estate agents Big Island" · "Buying a condo in Hawaii."

Market context cited: The Honolulu Board of REALTORS (HAR) March 2026 data — single-family home median price $1,199,500, condo median price $510,000 — is consistently surfaced when Claude is asked about Hawaii real estate market conditions. Submarket-specific brokerage citations (e.g., "best Realtor in Hawaii Kai") are sparse in the single-engine training-data baseline; submarket depth requires web augmentation. Detailed real estate marketing dynamics are in our Hawaii Real Estate Marketing Guide.

5.6 Healthcare (8 queries)

Query subset: "Best hospitals in Hawaii" · "Top healthcare provider Honolulu" · "Largest hospital in Hawaii" · "Pediatric hospital Hawaii" · "Cardiology Hawaii" · "Cancer treatment Hawaii" · "Mental health care Hawaii" · "Healthcare options on neighbor islands."

Patterns observed. Hospital systems dominate citations; specialty practices (specific cardiologists, dermatologists, mental-health providers) are largely absent from the single-engine training-data baseline. This is a real limitation of Volume 1 — for actionable patient-facing information at the practice level, web augmentation is required. The neighbor-island access pattern documented in our Hawaii Healthcare Marketing Guide shows up here too: Claude consistently routes specialty queries toward Oahu providers (Queens, HPH) even when the query specifies Maui or Big Island. Specialty practices on neighbor islands face a meaningful AI search visibility gap.

5.7 Weddings (7 queries)

Query subset: "Best wedding venues Hawaii" · "Best wedding planners Hawaii" · "Maui beach wedding venues" · "Where to get married in Kauai" · "Hawaii destination wedding photographer" · "Most beautiful Hawaii wedding venues" · "Hawaii honeymoon resorts."

Patterns observed. Wedding citations are venue-dominated rather than vendor-dominated in the single-engine training-data baseline. Specific planners, photographers, florists, and officiants surface unevenly and inconsistently across Volume 1 runs — this is a known limitation. The Hawaii Wedding Professionals Association (HWPA) member directory and the dominant national wedding directories (The Knot, WeddingWire, Junebug Weddings, Style Me Pretty) are the source layers that web-augmented runs in Volume 2 will pull from. Detailed wedding marketing dynamics are in our Hawaii Wedding Marketing Guide.

5.8 Home Services (5 queries — disclosed thin)

Query subset: "Best contractors in Honolulu" · "Hawaii landscaping companies" · "Pest control Hawaii" · "Best house cleaners Honolulu" · "Plumber Maui."

Volume 1 finding: home services is a citation desert in the single-engine training-data baseline. Specific named contractors, landscaping firms, pest-control companies, cleaning services, and trade providers do not surface reliably — Claude's training-data corpus has thin coverage of Hawaii home-services providers compared to hospitality, real estate, and healthcare, where consumer-facing brand authority (press, awards, reviews, hotel-directory presence) drives accumulated citation weight.

What does surface: generic guidance ("look for licensed contractors", "check Yelp reviews", "verify Hawaii contractor licensing through DCCA"). The Hawaii Better Business Bureau and the Hawaii Department of Commerce and Consumer Affairs (DCCA) Professional and Vocational Licensing Division are cited as the verification layer, but specific provider names are not.

Implication: Hawaii home-services businesses face a meaningful AI search visibility gap relative to hospitality and healthcare. This is also a meaningful opportunity — the category has structurally less competition for AI citation weight than the dense hospitality category. Volume 2 web augmentation will fill in specific provider citations from local directories (Yelp, Google Business Profile, Hawaii BBB).

5.9 Cross-category insights

Patterns that emerged across categories in Volume 1:

• Brand authority concentrates citations. Businesses with strong national brand affiliations (Disney, Four Seasons, Hilton, Berkshire Hathaway, Sotheby's) earn citation weight that transfers from mainland brand recognition into Hawaii-specific queries. Independent flagship properties without that affiliation (Halekulani, The Royal Hawaiian, Mama's Fish House) earn equivalent citation weight through accumulated Hawaii-specific press authority — a slower path, but durable once established.
• Multi-property and chef-anchored businesses cross-cite across queries. Halekulani as hotel and wedding venue; Roy and Merriman across multiple restaurant queries. Multi-touchpoint brand presence compounds.
• Geographic citation skew toward Oahu and Maui. Kauai, Big Island, Lanai, and Molokai appear primarily through their flagship properties (Princeville, Four Seasons Hualalai, Four Seasons Lanai, Mauna Lani, Pacific Whale Foundation). Submarket coverage on neighbor islands is materially thinner.
• Category citation density varies dramatically. Hospitality is dense (15+ named hotels, 12+ named restaurants surface readily). Home services is sparse to absent. Healthcare hospital systems are dense; specialty practices are sparse. Real estate is dense at the major-brokerage level; sparse at the individual-Realtor level.
• Hawaii Regional Cuisine chefs anchor restaurant citations. Roy Yamaguchi, Peter Merriman, Alan Wong, and Sam Choy collectively account for a meaningful share of Hawaii fine-dining citations through their named restaurants. Brand authority for HRC chefs propagates across queries beyond what individual restaurants would earn standalone.
• Neighbor-island specialty access gap is visible at the AI level. Healthcare specialty queries route disproportionately to Oahu providers even when neighbor islands are specified — the same gap documented in our healthcare pillar guide shows up directly in citation patterns.
• National wedding directories will be Volume 2's source layer. Wedding vendor visibility (planners, photographers) is materially under-represented in the training-data baseline because most vendor-level discovery happens through directories (The Knot, WeddingWire) and visual platforms (Pinterest, Instagram) that need web augmentation to surface in citation data.
• HAR market data is the most consistently cited Hawaii business statistic. The Honolulu Board of REALTORS' monthly market figures (median price, inventory, days on market) surface across real estate queries with high accuracy and recent timestamps, suggesting AI engines have weighted hicentral.com as the authoritative Oahu market source.

5.10 Volume 1 limitations and Volume 2 roadmap

Volume 1 has five primary limitations, each addressed in Volume 2:

1. Single engine. Anthropic Claude only. Volume 2 expands to ChatGPT, Perplexity, Google Gemini / AI Overviews, and Microsoft Copilot per the framework's documented engine list.
2. Single run per query. The framework specifies five runs per query for non-determinism handling. Volume 1 uses single-run sampling. Volume 2 applies the full five-run protocol.
3. Training-data baseline. Volume 1 reflects what Claude knows from its training corpus, without web augmentation. Recent business openings, closings, rebrands, or news cycles are not captured. Volume 2 uses web-augmented runs.
4. 50-query subset. The full methodology specifies 500–1,000+ queries per category. Volume 1 applies a 50-query universe across categories. Volume 2 expands toward the full universe.
5. Home services category thinness. Single-engine training-data coverage of Hawaii home-services providers is minimal. Volume 2 web augmentation will pull from local directories (Yelp Hawaii, Google Business Profile, Hawaii BBB).

Volume 2 publication target: Subsequent quarters, with timing dependent on the operational pipeline build for multi-engine, multi-run sampling. Volume 1 establishes the comparison baseline; Volume 2 will report quarter-over-quarter changes for businesses cited in both.

6. How Do You Apply This Framework to Your Business?

The framework is built as a public methodology. Three primary user groups apply it for distinct reasons.

Hawaii businesses applying it for self-assessment

Any Hawaii business can apply the framework to its own visibility. Volume 1 is the comparison baseline: see whether your business surfaced in the relevant category table; if not, run the framework's queries against multiple AI engines yourself, capture whether your business appears, and score against the rubric. Repeating the run quarterly produces a trend line. Most Hawaii businesses are doing this for the first time when they apply the framework — the visibility data simply has not existed before.

Researchers and journalists publishing comparative datasets

Researchers, journalists, and analysts can apply the framework to produce their own datasets and publish them with attribution under Creative Commons BY 4.0. Compare Volume 1 against your own measurements — same queries, different engines, different time periods, different scoring weights. The framework's value is exactly that multiple parties applying it produce comparable results.

Agencies (including Nekko Digital) applying it as part of client work

The framework is the methodology Nekko Digital uses inside our AI search optimization engagements when measuring client baseline AI visibility. Other agencies are encouraged to use it for the same purpose. The standardization makes client-facing AI visibility claims independently verifiable.

Refining the framework

AI search platforms and behaviors evolve. Methodology refinements are welcome and applied openly: any party proposing a methodological change is expected to publish the proposal, the rationale, and the impact on existing measurements. Refinements take effect prospectively — applied to runs conducted after the refinement is published, never retroactively to existing data — so historical datasets like Volume 1 remain interpretable under the rules in effect at the time of their production.

7. How This Fits With Our Other AI Resources

The Index is one of six AI search resources on the Nekko Digital site. Each serves a distinct intent.

• AI Search Optimization — Strategy hub. Explains the GEO and AEO disciplines and how they fit together at a strategic level.
• Generative Engine Optimization (GEO) — Service. The transactional side of GEO work.
• Answer Engine Optimization (AEO) — Service. The transactional side of AEO work.
• AI Search Playbook — Practical reference. Platform comparison tables, content frameworks, and a 30-day action plan.
• AI Search Self-Audit — Self-assessment. 41-point interactive audit of your own site.
• Hawaii AI Search Visibility Index (this page) — Open methodology framework + applied datasets. The published rubric for measuring AI citation rates of Hawaii businesses.

A Hawaii business doing serious AI search work will use most or all of these. The strategy hub explains the discipline. The Field Manual and Readiness Audit guide the implementation. The GEO and AEO service pages cover what we do in engagements. The Index provides the open standard for measuring whether any of the work is producing visibility.

8. FAQ