4.1 Buyer groups and annual opportunity value
What you'll do in this section: take your own clinic's price list and recent sales records, fit them into the five buyer groups below, work out the annual opportunity value, and use that number to rank which one to go after first. Also remember that the "get more reviews" tactic does not work for dental clinics; what you swap in instead is three other things.
Swap all five numbers in the figure for your own. The formula is annual opportunity value = price per order × gross margin × monthly capacity cap × 12; all four numbers are copied from your own price list and recent sales records, so anyone can recompute it. The full definition of this formula and the four rules below is in → General Edition 4.5 Buyer types, annual opportunity value and the three checks; for dental clinics you only change the numbers here, never the calculation.
Copy these four rules down together with the numbers:
- Fill in the numbers yourself, confirm them in person, and put them somewhere visible in the plan.
- The monthly capacity cap is a hard cap—if you can't do 8 implants a month, don't pick a question built for 8.
- Annual opportunity value is only used to rank; never describe it externally, in any wording, as a revenue forecast. Never multiply it by, divide it by, or show it side by side with seat counts or enquiry counts.
- When two questions' annual opportunity values are close, pick the one with less competition.
Buyer groups like full-mouth reconstruction, which close by phone and are hard to attribute, are not dropped from the pool just because attribution is hard: whether you can attribute is a site-level capability, not a property of the question. From the day work starts, it is accepted on the "observable tier"; how the two tiers are split is in the three checks of → General Edition 4.5 Buyer types, annual opportunity value and the three checks.
Don't bolt on a separate "score four factors 1–5 each, then multiply" table. In that kind of scoring, three of the factors actually measure the same thing, and the fourth is essentially ranking by search volume, which systematically pushes dental's low-frequency, high-price-per-order items (full-mouth reconstruction, for example) to the bottom.
Dental and aesthetics buyers are consumer buyers. Cross-industry research finds that consumer services win on review volume (sample did not include Singapore, 75% confidence; evidence in → General Edition A.2 Evidence for picking targets, writing pages, off-site and retests), but for dental clinics that tactic cannot be turned into "get more patients to write reviews". What actually moves these buyer groups is three things: price pages, device and process facts, and how complete the five business profiles are (price wording is in chapter 1, 1.2; business profiles are in chapter 3, 3.4). For dental clinics, reviews are only passively observed, never chased as an active lever (see chapter 6, 6.5).
4.2 Four states and the ten question types: where dental questions go
What you'll do in this section: take every URL in the intent-level top 10 and judge each one against the four states; then match the ten common types of dental question to the action each should get, with second opinion placed in the first batch. When a policy question is filled up by government sites or big platforms, rephrase it and judge it once more; don't write it off outright.
How to judge the four states, which four thresholds "reachable enough" must clear, and how the abstain line is calculated are fully defined in → General Edition 4.6 Four states, two denominators and the abstain line; for dental clinics you add three specific values:
- In the five "not reachable" classes, registers and notice pages maintained by government or statutory bodies are, for dental clinics, the MOH HCI Directory (the Ministry of Health's directory of licensed healthcare institutions) and hpp.moh.gov.sg. Count them as not reachable directly, by class.
- Of the four "reachable enough" thresholds, the "allowed on this side" one: Best / Top list pages, even when they look open to submissions, do not count as reachable enough for dental clinics. Do not submit to these pages and do not pay for a listing; if the clinic is already on one, ask the publisher in writing to remove it (chapter 6, 6.2).
- When the four-state judgement lands on the "reachable · dominated by lists / directories" cell, first run it through the separate rules for paid listings and for wording in chapter 6, 6.2, and only then decide whether to submit or pay for a listing.
Once the four states are judged, assign an action by question type:
Types the figure doesn't show: pure symptom questions (such as "what causes swollen gums") and plain educational questions are already screened out at the question-pool step (chapter 2, 2.2) and never reach this point; scenario questions ("does this suit my case") are done once they pass the three checks.
Rephrase policy questions first, then judge them again. A question like "how much can Medisave cover for a dental implant", filled up by government sites or big platforms, must not be counted straight against the abstain line as "not reachable". First turn it into an execution question such as "if you have it done with us, can you use Medisave, and do we help you claim it", then run it through the four states again; only if it is still locked out after the rewrite does it count against the abstain line. Writing it off outright would write off the entire Singapore dental and aesthetics industry (verified by live searches). The rewritten execution question goes to ⑥ one question, one page in chapter 5, 5.9, and that page carries only the regulation section (the official figures get a section of their own, and the clinic's own price never appears inside that section; see chapter 1, 1.3). The clinic's own price stays on the ① single-service price page, and this page only carries one link across to it. Do not write a sentence like "how much you pay after Medisave is deducted" that folds the subsidy into the clinic's own price.
Treat "how much" questions and symptom questions that carry an option and a price as price pages—price-type questions trigger AI Overviews at the highest rate of any question type, and they are one of the few cells where your own site can actually make it into the answer (US sample; do not write it as a Singapore figure; evidence in → General Edition A.2 Evidence for picking targets, writing pages, off-site and retests). A symptom question that carries an option and a price, such as "my gums are swollen: do I need an extraction, and how much does it cost", is a shelf question. Treat it as a price page; don't drop it from the pool as a pure symptom question. Always write prices as a fixed final price; for wording, follow chapter 1, 1.2.
"Which is better, A or B" compares procedures and treatment options only, never clinics—across all industries, comparison pages have the highest citation density of any page type, and that is counted by page type, not by domain, so the main battlefield is "is there a qualifying comparison page under this question", and whose domain the page sits on is a secondary question (evidence in → General Edition A.2 Evidence for picking targets, writing pages, off-site and retests). Never compare clinics, never compare peers — that is a red line; for wording, see chapter 5, 5.7.
"Which clinic do you recommend" works in two steps: off-site gets you on the shortlist; your doctor pages and procedure pages get you through the check—AI first searches out a shortlist of clinics, then checks them one by one (the question breakdown is at the end of this section). On 2026-09-29 we put 8 dental and aesthetics services to ChatGPT: of the clinics named when the question was a generic "best", 50/71 had their own website cited in the same answer; when the question was "best" plus a specific condition or patient group, 76% of the clinic pages cited were deep pages such as doctor pages and procedure pages. This is correlation, not causation — AI may also have picked the people first by registered specialty and reputation, then gone to their websites for material to check them against (evidence in → General Edition A.2 Evidence for picking targets, writing pages, off-site and retests). At the shortlist step, 36/62 of the names on generic "best" lists overlap with third-party best-of lists; but for dental clinics, third-party Best / Top pages are never for seeking a listing or paying, and if the clinic is already listed you must also ask the publisher to remove it (chapter 6, 6.2). So what you can do off-site is pure directory listings, Group B sources and the five business profiles. Questions phrased as "which clinic nearby" mostly fall into the maps / reviews cell, and what needs filling in is the fact fields on your business profiles (chapter 3, 3.4), not another article.
"Best" plus a specific condition or patient group can take one slot as a trial; don't make it a main target yet—for example, "which orthodontist should an adult with a deep bite see". Questions phrased this way produced a list in every answer (ChatGPT 16/16, AI Mode 8/8); the list named a different set of clinics from a generic "best", and asking the same question twice gave a steadier list (list overlap 0.46, against 0.28 for a generic "best"); in the answers, 56/65 of the named clinics were described as specialists (judged from the answer text, not checked against the register). But no intervention experiment has yet shown whether filling in doctor pages and procedure pages gets you onto the list, and there is no figure for how many people ask this way. So give it at most one slot this quarter: complete the register fields on the doctor pages and the matching procedure pages, retest on day 30, and promote it to a main slot only once you are on the list; a clinic without a matching registered specialist does not schedule this slot yet (inferred, 60% confidence).
Second opinion / corrective follow-up goes into the first batch: highest price per order (dental: S$8k–18k), least price-sensitive, and patients are bound to search for it on their own. If this group gets stuck in the cell where "the page is cited but no one is named in the answer" — that is "no one has been named in the answer yet", not "no one is doing this" — adding pages doesn't help; what needs fixing is how the sentences are written: checkable numbers, no adjectives, a source. In the build, first enter the pages already competing for this question's citation slots into the citation-slot table; don't just write the one page on your own website.
This wording rule applies the same way in the H1, H2, body text, title and off-site materials; any wording pointing at another clinic or doctor is banned, named or unnamed. This group's name is "second opinion / corrective follow-up clients"; the label "failed-case repair clients" must not appear in any external document. The page's clinical skeleton is in chapter 5, 5.11.
When talking externally about this type of question, you may say only one sentence: "Who AI names for these questions today, and whose pages it cites as sources, we will count on the spot in the baseline." ❌ Never say "no one is doing this" or "competition is near zero", and never report a specific number of businesses. That number comes from a generic search with the region not locked, on a single engine, run once, and it drifts over time: if the other side searches for themselves next week and sees nine businesses, you lose their trust just as you would by saying "competition is near zero".
Brand questions (is X trustworthy, are there complaints, did AI get our facts wrong) go into the billing layer, packaged as a standalone deliverable—this is the only one of the ten types that "slips back if you stop", and the one cell you can see with your own eyes and reproduce yourself: once a fact is fixed, what AI says changes to match, and the monthly retest shows it. Give it away as a freebie, and you give away for nothing the one thing that really works.
Question breakdown: the pages a dental clinic needs
The method follows the "question breakdown" in → General Edition 4.7 Actions by state, the ten question types and this quarter's slots. The table below covers 74 ChatGPT answers for dental and aesthetics (64 from 2026-09-29, plus 10 from 2026-09-14) and counts, for each type of question, how many answers checked each category:
| What AI checks | Which questions check it most | Which page answers it |
|---|---|---|
| Specialist registers and official sources (SDC / SMC / MOH) | generic "which clinic is best" 21/22, specific "which clinic is best" 14/16, price questions 13/18, detail questions 15/18 | Register fields on the doctor page (3.2) + verification and lookup page (5.17); the registers themselves you only match, never change |
| Specialist terms (the patient's words turned into procedure names) | generic 15/22, specific 15/16, detail 11/18 | Eligibility and process pages written in specialist terms (5.10), definition pages (5.12), criteria-based selection guides (5.8) |
| Price and subsidies | price 18/18, generic 8/22 | Price guide page (5.4), single-service price pages (5.5), subsidy page (5.6), Medisave one question, one page (5.9) |
| Safety, guidelines and literature | detail 14/18, specific 11/16 | Regulatory document alignment page (5.15), eligibility and process pages (5.10); this cell is mostly taken by government sources and the literature |
| Reputation | generic 20/22, specific 6/16 | The five business profiles (3.4); reviews are only passively observed (6.5) |
| Checking each named clinic (by clinic name or site: website) | generic 21/22, specific 13/16 | The named clinic's /facts (3.1), doctor pages (3.2), single-service price pages (5.5), eligibility and process pages (5.10), store / branch pages (5.16) |
The most widely useful page is the register fields on the doctor page: under all four types of question, more than seven in ten answers went to check official sources or registers. Next come procedure pages written in specialist terms. The finding-candidates step happens off-site: pure directory listings (the exempt directories in 6.2), Group B sources (6.3) and the five business profiles (3.4), never Best / Top lists.
Once all ten question types have an action assigned, you put them together into a list you can start work on this quarter with three combination rules: at least one hold-pile question, at least one brand / correction question, and at most one national head term with no location. Apply these three rules as set out in → General Edition 4.7 Actions by state, the ten question types and this quarter's slots; dental clinics need no separate values here. Once the slots are set, the first batch's build order is in chapter 0, 0.1.
