# Appendix C · Five industries side by side

> GEO Playbook · General Edition v1.0 · Canlah AI · CC BY 4.0 · Web page: https://canlah.ai/playbook/industry-reference/
> Markdown edition for AI assistants, same content as the web page. Figures are code blocks (wireframe / mermaid / bars / steps / split); "→" links open the matching section on the web, and the same URL with .md is its Markdown edition.

## C.1 Ten-axis values for the five industries covered, and three things for week one

**What you'll do in this section**: When you're judging a new industry, first find the industry in the table below whose ten-axis values are closest to yours and use it as a reference, then go back to the main text and follow the rules. These are only examples: do not copy another industry's tactics straight onto your own.

```steps Figure: This table is only a reference for step 4; tactics are still decided by the main text's decision table
① | The ten questions | 8.1 the ten questions — which axis each one fills in
② | Decide the side | 8.2 the five steps that set the side and switches, then back to 0.2 to double-check
③ | Fill in the ten axes | 8.3–8.4 fill in your own value for each axis
④ | Find a reference | pick the column below whose ten-axis values look most like yours
⑤ | Back to the decision table | 8.5 rule on each axis; do not copy that column's tactics
```

Every cell in the table is a summary, not a work order: the axes are defined in （→ 通用版 8.3 十个轴（一）：页型、容器、锚点、监管、问法） and （→ 通用版 8.4 十个轴（二）：决策人数、产能、归因、半衰期、语言）; the seven tactics are defined in （→ 通用版 8.6 七项弹药：受监管与不受监管差在哪、凭什么动名次）.

**Quick examples only; the rules are set by the main text.**

### Values on the ten axes

| Axis | Dental/aesthetics | Family law | Education and training | B2B software/SaaS | E-commerce |
|---|---|---|---|---|---|
| 1 Shelf page-type mix | News and long-form articles dominate (`article` 54% / `listicle` 0%) | Government, court and legislation pages dominate the ChatGPT leg (family law 47/60, 49/59); on the AI Mode leg, law firms' own explainer articles and fee pages lead (law-firm sites 28/52); rankings, review platforms and the Law Society directory together make up 10/171 and lead only on "who's best" questions | List pages dominate, with long-form price guides second | List pages + comparison pages together ≥60%; documentation sites about 8%; news about 1% | A mix of list pages and comparison pages. For platforms, read this as two separate sentences, not one: ① a platform's product detail pages feed only the shopping shelf; ② a platform's shop bio, category pages and product description text can be copied by the retrieval leg into the body of an answer (provided that platform's door is open on that leg — decide leg by leg). Sentence ② is the shortest link in path ① |
| 2 Control of citation containers | Mostly third-party slots; you can also take a few own-site slots | Read the two legs separately: the government and court pages the ChatGPT leg cites are not reachable; on the AI Mode leg, own-site slots can be taken (law-firm sites 28/52); directories and rankings are third-party slots and lead only on "who's best" questions | Mostly third-party and platform slots; you can also take own-site slots; 6–8 reachable-enough URLs | The most own-site slots (highest of the five industries); a whole row of platform slots; very few not reachable | Both own-site slots and platform slots are high; whether a platform slot can be used varies hugely from platform to platform — you must verify each with a live curl test |
| 3 Entity anchor layer | A person (the doctor); the organisation second | A person (the lawyer) first, the firm's approved name second (a judgement call, not measured) | Splits by sub-segment: organisation / course / organisation + instructor / person | Product SKU; the founder and engineers second | Dual anchors: parent brand (answer shelf) + SKU (shopping shelf) |
| 4 Side and switches | Strictly regulated side + switch A (agency liability) + switch E (peer comparison banned), already part of the strict side's base layer; add switch C (regulated products) when aesthetic devices are used, or when skincare or supplements are also sold | Lightly regulated side (75% confidence) + switch A (agency liability; the "principal bears it alone" form) + switch B (legally required fields; partly on: only fixed wording that specific conditions trigger) + switch E (peer comparison banned) + switch F (referral commissions banned; any form of consideration counts); C off, D depends on the client; family cases add the "three family-law rules", which tighten how the case-study, testimonial and review-reposting, and review-reply cells are written, without changing the side | Regulated: build the whole site to the strictest segment's standard (switch D (spans two sides)) + switch A (the rule text names the party publishing on the client's behalf) + switch B (ads must carry the registered name, registration number and validity period); whether it falls on the strictly regulated or the lightly regulated side is still to be verified (see below the table) | Unregulated side, with three pass-through red lines: certification wording is a legal claim, the regulation of your buyers' industry passes through to you, and platform terms are enforced faster than the law | Unregulated side; five sub-categories, including supplements and devices, turn on switch C; if any one cluster is hit, build the whole site to that cluster's regulated basis (switch D); switch A is still to be ruled on (see below the table) |
| 5 Question-shape mix | High share of spec type (price) | Spec type + recommendation type (the mix we set when writing the questions; not measured, must be recalculated against the frozen question pool) | Recommendation 4 / spec 3 / comparison 2 / approach 1 (approximate, not measured, must be recalculated against the frozen question pool) | Recommendation 35 / comparison 30 / spec 25 / approach 10 — the only industry where all four fronts must be fought at once | Filled in only after the frozen question pool is signed off; left blank until then — pre-filling is not allowed. Prior assumption (for sequencing the work only; it goes into no deliverable and is never said externally): the share of shopping-type questions is significantly higher than for training, and this must be measured against the frozen questions; do not copy the training column over and treat it as your conclusion |
| 6 Number of decision-makers | Single, over weeks | Depends on the matter (a judgement call, not measured): a divorce is one person (the other party is the opponent); wills and LPAs are often a couple | K12: two people + a third-party user (the child); corporate training: committee | Committee (user / IT and security / finance / legal) | Single, on the spot; for high-ticket durable goods, single over weeks or two people |
| 7 Hard capacity cap / work unit | Practitioner-chair count / one patient type | Lawyer hours / one client type | Class capacity × how often classes run / one course line | No hard cap → switch the numerator / one use case × one role (this is the pricing unit; pages are built one per use case, at most 4 use-case pages a quarter) | Inventory and supply volume / one category cluster |
| 8 Attribution channels | Phone + in-person visits + forms | Forms + phone + WhatsApp (a judgement call, not measured) | WhatsApp (the primary channel) + trial-lesson forms + campus visits | Free-trial sign-up / demo / docs site → sign-up | Add-to-cart and checkout (on platforms there is almost certainly no referrer) |
| 9 Content half-life | Medium/long (driven by procedures and regulation) | Long (driven by regulation) | A mix of three levels, decided field by field | Short (quarterly); compliance evidence: medium; docs site: event-driven | A mix of three levels, decided layer by layer: ① the live-numbers layer, ≤15 minutes — price, stock and promotions, managed by the feed and never entering the body text; ② the spec-and-ranking layer, monthly — spec numbers, list-page ranking, the values in comparison tables; ③ the structure-and-basis layer, quarterly — price ranges, definitions, the variables that decide price, and how buyers choose. The review cadence follows the layer, not the page |
| 10 Language fork | The difference set often holds (medical tourism) | Depends | The difference set holds, with the cleanest evidence | No language fork, but a jurisdiction fork holds | Decided by the buyer mix; for the Singapore market, run the Chinese difference set first by default |

### Tactics that follow from the ten axes

| Tactic | Dental/aesthetics | Family law | Education and training | B2B software/SaaS | E-commerce |
|---|---|---|---|---|---|
| Main battlefield | Off-site + own-built price pages | Own-built fee pages and explainer articles (the AI Mode leg mainly cites firms' own pages); off-site directories and rankings only for "who's best" questions | Platform slots + own-built tuition-fee pages + list-page editorial slots | Own-site pages (the highest of the whole table) | Fighting on both shelves at once: own-built media-style pages + the feed |
| What to build first | Tiered fixed-price page | Tiered fees and billing-method page | All-category tuition-fee guide page (English and Chinese published together) | `/pricing`, eight blocks | An all-category comparison page that includes competitors |
| Can you get onto "Best" lists | Do not do it; if you are listed, ask the publisher to remove you (labelling it as paid is not a release from liability) | Self-nomination not allowed; can take part via editorial slots | Possible in some segments (editorial slots); in the strictest segment, entry descriptions must not be laudatory | The main tactic | Can get into editorial review slots; cannot buy a slot |
| Asking for reviews | Never ask; passive observation only | May ask, on four conditions: no incentives, no ghost-writing, no cherry-picking (no gating), no dictated content; the basis is that no current rule was found banning it, not that a rule allows it (70% confidence); family clients' reviews are not reposted by default | Can do it, split by campus | Mandatory, but the platform's terms are the gatekeeper | Mandatory, with three hard boundaries: ① no gating; ② any incentive must be disclosed, and applied the same way to every invited buyer; ③ never have staff, friends and family, or fake-order accounts write reviews, and every review you display must be visible text |
| Competitor comparison | Compare procedures only, never clinics | Do not do it | Compare only checkable, objective items | The main tactic | The biggest tactic here, compared with the industries before it |
| Weight of earned news | High (`article` 54%) | High | Medium | About 1% (off-site work ranked P3) | Medium (review sites and media listicles are the main source, but not news) |
| Number of retest rounds | 3 | 3 | 3 | 5 | 3 (but each of the two seat rulers measures its own noise band) |
| Special door pitfalls | WAF | None | Timetables/price lists/faculty lists rendered in JS | Multiple subdomains + docs-site CSR + hosted trust pages | CSR + numbers with no sentences + the platform's own door |
| The only change to the ruler | None | None | Split the denominator by grade level × subject | Split factual errors into two columns; measure total names by category | Count seats with two separate rulers; add price errors and stock errors to factual errors |
| Net advantage / net disadvantage | Advantage: the official registration number gives you the strongest anchor for free | Advantage: the official register lookup (the LSRA lawyer and law-practice search) gives you an anchor for free; whether it has a publicly citable number field is still to be checked | Disadvantage: instructor turnover breaks path ③ (getting identified correctly) | Disadvantage: namesake-entity pollution; Advantage: the most own-site slots | Disadvantage: one extra door layer that is not in your hands; Advantage: product data supplied directly (the feed) never goes through the door |

The four kinds of container in the "control of citation containers" row — own-site slots (you can claim these on your own website), third-party slots (need someone else's agreement), platform slots (fill in the back end and they take effect), and not reachable — are defined in 8.3, axis 2; what the six switches in the "side and switches" row each mean is in 0.2.

The "side and switches" row is decided using the four questions and six switches from （→ 通用版 0.2 先判你属于哪一侧）. Dental/aesthetics and family law have published industry editions, with the rule text and sources in each book's appendix; the other three columns are reference results from the four questions — the rule text has not been checked clause by clause, so this is not a final ruling. The family-law cell is now decided; two cells are still unresolved:

- **Family law**: decided as the lightly regulated side, 75% confidence. Switches A (agency liability), E (peer comparison banned) and F (referral commissions banned) are on, B (legally required fields) is partly on, C is off, and D depends on the client; family cases add the "three family-law rules", which only tighten how the case-study, testimonial and review-reposting, and review-reply cells are written, without changing the side. For the basis, the strongest counter-argument and the conditions for changing the ruling, see Chapter 0 of the Law Firms Edition.
- **Education and training**: for the strictest segment (registered private education institutions), testimonials and objective comparisons are both allowed with limits, which points to the light side under question 3; but for tuition fees it requires eight items, each written as a fixed figure, and the original text does not say whether a range can be written (question 2). Which side it falls on is still to be verified in the training edition; until then, write to the stricter, strictly regulated side.
- **E-commerce**: the original text contradicts itself on the agency-publishing cell. One place says there is no agency liability and that set of authorisation paperwork is not needed; another says no rule has been found that names the publishing party, yet every batch must still be signed off. This is still to be ruled on by the industry edition; until then, follow the stricter practice and sign off every batch of external material.

Even if your industry is one of the other three, still check your own industry's original rules against （→ 通用版 8.2 判侧五步与三条判据） before you decide the side; when unsure, start on the stricter side.

On penalties for agency liability, the original text gives figures for two industries: dental/aesthetics — a fine of up to S$20,000, up to 12 months' imprisonment, or both, plus up to S$1,000 a day for a continuing offence (checked against the original rule text); for the strictest segment of training, a fine of up to S$5,000 or imprisonment of up to 6 months for false or misleading advertising, and a fine of up to S$10,000 or imprisonment of up to 12 months for not complying with an order to withdraw and correct. Rule numbers and quotations from the original text are not included in this appendix: for dental/aesthetics, see *Dental & Aesthetics Edition*, Appendix A · Rule text and sources; for family law, see *Law Firms Edition*, Appendix A; for the other industries, these will be added once their industry editions are published.

Cells in the table that say "Do not do it" or "Never ask" do not all rest on the same label tier; the clause-by-clause labels in the relevant industry edition decide; where something is banned as Statute text, a client's written sign-off cannot make it publishable — a written sign-off resolves only the three stop-and-escalate situations; it never turns something banned as Statute text into something publishable (see （→ 通用版 0.4 合规句的三档标签与停笔规则）).

### Tactics ruled out and limited

"Ruled out" means that tactic cannot be used at all; "limited" means it can be used but with hard boundaries, not a blanket ban. The table only fills in figures the original text gives; where none is given, it says so — nothing is estimated.

| Industry | Ruled out | Limited |
|---|---|---|
| Dental/aesthetics | Four tactics | Not listed separately in the original |
| Family law | Two and a half tactics | Not listed separately in the original |
| Education and training | Spans tiers; the original gives no figure | Spans tiers; the original gives no figure |
| B2B software/SaaS | Zero tactics | Zero tactics (plus three pass-through red lines — see the "side and switches" row above) |
| E-commerce | Zero tactics | Three tactics |

### Three things for week one (examples)

Examples only: reorder the three things your industry does in week one to match the side-deciding and baseline order in （→ 通用版 0.1 全书一句话与 90 天翻书顺序）; do not copy the table below directly. For the first item in the table, "the five door gates", week one is a read-only audit only; fixing the door has to wait until after the baseline is saved.

| Industry | First | Second | Third |
|---|---|---|---|
| Dental/aesthetics | The five door gates + `/facts` | Tiered itemised fixed-price page | The five business profiles + checking official registers + manufacturer locators |
| Family law | The five door gates + one page per lawyer | Government and association registers, without missing a single one | Tiered fees and billing-method page |
| Education and training | The five door gates (including a CSR check on the timetable) | Three platform-slot (C) items: course-catalogue fields + the five profiles for each campus + platform instructor pages | Tuition-fee guide page, English and Chinese launched on the same day |
| B2B software/SaaS | The five door gates + four special gates | The eight blocks of the pricing page + removing the `nosnippet` tag | A visible HTML mirror of the Trust page on the main domain + checking for namesake conflicts |
| E-commerce | The five door gates + three tables for CSR / visible text / platform robots (decide leg by leg, not the platform's overall state) | Aligning the dual entity anchors: ruling on the brand name vs the model name + rewriting the shop bio on each platform | Submitting the feed application + a 100-item sample batch (does not go through the door). Week one is the submission and the sample batch, not activation: activation has to clear the other side's business verification, about 1–3 weeks; the feed must not be pushed until the dual anchors are aligned, and the full push is scheduled after the second item |