What a $4,500 kickoff actually produces.
An anonymized example of the research memo and prepped punch list a paying customer receives at the end of the three-week kickoff engagement.
1. Executive summary
For the questions a Honolulu med-spa shopper actually types into AI, the engines name the same handful of firms over and over. One competitor appears on all 18 measured questions. Four more appear on the majority. If your firm is not in that set, you are absent on exactly the demand-shape queries that bring in new patients, while named competitors are present consistently.
Where the contest lives is specific. The five web-searching engines cite Honolulu med spas' own websites heavily (Gemini 64%, Perplexity 61%, ChatGPT search 57%, Google AI Overviews 53%), so on the surfaces where most AI discovery happens, your own site does more than half the work of getting cited. Two engines are category-wide blind spots: Claude cites med-spa sites just 2% of the time and Microsoft Copilot 0%. Those two are not a you-problem, they are an opening that sits open for every firm at once.
The punch list at the end of this memo orders the actionable work by impact. Item 1 is the highest-leverage single piece of work, and it is the same page shape competitor A has on every question it is cited for.
2. Methodology and scope
Eighteen hash-locked questions a Honolulu med-spa buyer would actually ask, split across head-intent ("best med spa in Honolulu", "top rated medical spa Honolulu"), service-intent (Botox, filler, laser hair removal, CoolSculpting), neighborhood-intent (Waikiki, Kakaako, Oahu), and trust-signal queries. Each question ran three times per engine per run to separate signal from noise. Question set locked at hash 26851ed8 so every future run compares cleanly.
Engines measured (all 7 from day one)
- Five citation-grade engines: Perplexity, ChatGPT search, Gemini (grounded with live web search), Microsoft Copilot via Bing, Google AI Overviews.
- Two model-knowledge engines: Claude and Gemma (open-weight), which tell you what AI says about the category when it cannot search the live web.
Cohort
The 15-firm cohort was built from the citations themselves, not picked in advance: a first run with no cohort registered, then a coverage scan that surfaced every firm the engines cited heavily for these questions. That is why the competitive map is the set of firms AI actually names, not a list we assumed.
What this engagement did not measure
- Voice AI surfaces (Siri, Alexa, Google Assistant). Out of scope until measurable through public APIs.
- AI ad placements. Out of scope until those formats are measurable.
- Geo variation outside Oahu. Citation patterns on other islands may differ.
- Long-term causation. We measured what AI cites. We did not test whether shipping the punch list causes citation movement. The monthly delta memo reports observed movement honestly.
3. The headline finding
For every demand-shape query (a shopper asking AI to find a provider in your category without naming you), the engines return the same short list of competitors. The firms they named most often, across all three runs and all seven tools:
- Med Spa A: cited 531 times, present on all 18 of 18 questions, across 6 of 7 engines
- Med Spa B: 307 citations, 15 of 18 questions
- Med Spa C: 307 citations, 14 of 18 questions
- Med Spa D: 189 citations, 14 of 18 questions, the widest engine reach after A (5 of 7)
- Med Spa E: 172 citations, 15 of 18 questions
Med Spa A is not the largest med spa on Oahu by chair count or revenue. It is the firm whose authority signals the engines have learned to trust for this category. That distinction, the gap between market size and AI-citation share, is the entire work of the engagement.
4. Source-type distribution
Across every citation captured for the category, this is where the AI engines pulled from:
| Source type | % of citations | What it means for you |
|---|---|---|
| Independent web (third-party content) | 51% | Editorial pages, vendor and blog content about med spas |
| Competitor (med-spa-owned websites) | 45% | The firms' own sites. This is the share you compete for directly. |
| Review directories (Yelp, RealSelf, Healthgrades, Vagaro) | 3% | Low. Med spa is a firm-heavy category, not a directory-driven one. |
| Wikipedia | 1% | Minor |
| YouTube | 1% | Minor |
| Reddit, social, forum (combined) | 0% | Not a citation source for this category |
The non-obvious finding. Review directories are only 3% of citations for Honolulu med spas, far lower than a category like dental, where insurance-carrier directories dominate. The engines mostly choose between med spas' own sites (45%) and third-party editorial content (51%), not aggregators. The practical read: directory-claiming is low-leverage here, and own-site work is where the citations actually move. That is a real reversal from categories where directories carry the weight, and it is the kind of category-specific finding the kickoff is paid to produce.
5. Per-engine breakdown (all 7 engines)
How often each engine cited a Honolulu med spa's own website, versus third-party content. The spread across engines is the finding a single-engine tool would miss entirely.
| AI tool | Med-spa-owned share | Third-party share | Total mentions |
|---|---|---|---|
| Gemini grounded | 64% | 35% | 1,590 |
| Perplexity | 61% | 37% | 1,194 |
| ChatGPT search | 57% | 43% | 895 |
| Google AI Overviews | 53% | 42% | 172 |
| Gemma (training data) | 32% | 68% | 307 |
| Claude (training data) | 2% | 98% | 583 |
| Microsoft Copilot (Bing) | 0% | 92% | 802 |
Two rows decide the strategy. The top four (the web-searching engines) cite med-spa websites 53 to 64% of the time, so on the surfaces where most AI discovery happens, your own site is doing more than half the work. The bottom two are the blind spots: Claude at 2% and Microsoft Copilot at 0%.
The Claude blind spot, and why it is not yours to fix alone
Claude answers from training data, not live search. At 2% own-site share, it has effectively no memory of Honolulu med spas. We measured the identical collapse for Hawaii CPA firms (Claude 1%) and Austin CPA firms (Claude 0%): two unrelated local-service categories, same near-zero result. It is a property of how thinly local-service businesses appear in the editorial text a model learns from, not a quirk of one firm. The honest read: Claude is a category-wide blind spot no single med spa closes on a short timescale. Gemma (32%) is more reachable, through sustained editorial presence over the training cycle.
The Microsoft Copilot opening
Microsoft Copilot answers using Bing's organic results. Across 802 mentions for these questions, zero went to any med spa's own website, because the current Bing top results for these queries are dominated by directories and editorial content rather than individual firms' sites. The opening is open to everyone at once: in our data, the Honolulu med spa ranking first in Bing organic for the common queries is the one Copilot tends to surface, while competitors absent from Bing organic are absent there too. We name the condition. Whether changing it moves the citation is a measurement question we keep answering month over month.
6. Competitive gap analysis
For the demand-shape questions where you do not appear, the table below shows which competitors did, by total citations, how many of the 18 questions they reached, and how many of the 7 engines named them.
| Competitor | Citations | Questions (of 18) | Engines (of 7) |
|---|---|---|---|
| Med Spa A | 531 | 18 | 6 |
| Med Spa B | 307 | 15 | 4 |
| Med Spa C | 307 | 14 | 4 |
| Med Spa D | 189 | 14 | 5 |
| Med Spa E | 172 | 15 | 3 |
| Med Spa F | 105 | 12 | 4 |
| Med Spa G | 102 | 10 | 3 |
| Med Spa H | 87 | 11 | 3 |
| Med Spa I | 56 | 4 | 4 |
| Med Spa J | 45 | 5 | 3 |
Notable structural observations:
- Med Spa A is in a tier of its own: present on every question and across six of the seven engines. Closing the full gap to A is a multi-quarter authority project, not a quick fix.
- Med Spa B through E form the contested middle, each on 14 to 15 of 18 questions. This is the realistic target tier, and it is where focused punch-list work tends to have the shortest observable-movement windows. We do not promise a timeline.
- Med Spa F through J show up on 4 to 12 questions each, often on narrower service or neighborhood intent. These are where focused effort produces the fastest, most measurable movement.
- Engine reach is its own signal. Med Spa D reaches five engines on fewer citations than B and C, which reach four. Breadth across engines, not raw citation count, is often the more durable position.
7. Prepped punch list, ordered by impact
Sequenced by our estimate of leverage relative to implementation cost. Item 1 is the highest-leverage single piece of work on the list. This is a prioritized punch list you or your agency execute. Note that the work is concentrated on your own website, because the data shows that is where med-spa citations actually live.
Priority 1 · Build out service-by-neighborhood pages on your own site
Why: The web-searching engines cite med spas' own sites 53 to 64% of the time, and the firms that reach the most questions (A through E) all have own-site pages that name the specific service and the Oahu neighborhood served. For a category where directories are only 3% of citations, this is where the citations are won.
What: A page per high-value service (Botox, filler, laser hair removal, CoolSculpting) that explicitly states the neighborhoods served, the service variants offered, and who the service is for. The real deliverable lists which exact service-and-neighborhood combinations your competitors own and you do not.
Priority 2 · Show up in Bing organic, where Microsoft Copilot pulls its answers
Why: Copilot cites med spas' own sites 0% of the time because Bing's organic top results for these queries are directories and editorial content. In our data, the firm ranking first in Bing organic is the one Copilot tends to surface, while competitors absent from Bing organic are absent there too.
What: Verify the site in Bing Webmaster Tools, confirm the high-value service pages are indexed there, and add complete structured data (MedicalBusiness or LocalBusiness plus Service schema) so Bing can read the site as the provider for these queries.
Priority 3 · Audit authority backlinks for currency
Why: The engines weight authority backlinks heavily for this category, and stale, broken, or redirect-chained links signal lower trust. We checked your inbound link profile against the leading competitors and flag the links that 404 or redirect through chains.
Priority 4 · Earn editorial mentions to reach Gemma over the training cycle
Why: Gemma cites med-spa-owned sites 32% of the time, the highest of the two training-data engines, and it rewards genuine editorial presence rather than live-search ranking. Sustained third-party coverage is what moves it.
Priority 5 · Refresh meta descriptions on the high-value service pages
Why: The web-searching engines read meta descriptions as one signal of page topic. Your meta descriptions on the highest-value service pages are either missing or generic.
Items 6 through 12 (lower priority) are listed in the appendix of the customer deliverable, with the same effort and window estimates.
8. What happens next (in a real engagement)
- Day 1 of ongoing: measurement across the seven AI surfaces continues on a fixed cadence. A brief status note confirms each run completed.
- End of month 1: the first monthly delta memo. What moved, what did not, why we think it moved, and any punch-list updates.
- Ongoing: the punch list refreshes as your team ships items and as the engines shift. Drift detection flags category-wide pattern changes that affect your numbers.
How to scope this for your category
Email Lance@hi.neverranked.com with the category you want to measure and three to five competitors you would want on the cohort. The free one-page diagnostic runs five real buyer questions for your category across all seven AI tools and shows you which competitor firms AI is currently naming. One per business.
Full engagement is $4,500 to kick off a category and $1,500 a month after for ongoing measurement. Per category, not per client. Month to month after the kickoff.
Anonymized example deliverable. Category data measured 2026-06-09. Question set hash 26851ed8. The method is documented at /methodology/, every figure traces to a dated run on the /claims/ ledger, and the question set is hash-locked so future runs compare cleanly.