# Chapter 5 · Writing pages (2): the 9 most-cited page types, block by block > GEO Playbook · General Edition v1.0 · Canlah AI · CC BY 4.0 · Web page: https://canlah.ai/playbook/page-types-core/ > 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. Nine page types, each measured on both legs. One card per type: Intent / which leg / title and H1 / blueprint / standard column headers / sample sentences / most common mistake. Compliance gets only two lines: availability by side and cell-by-cell wording are both in 5.1–5.3, so here we add only the one line specific to this page type. Cited samples, n =, where the copied sentence sits, and length statistics are all in （→ 通用版 A.3 页型实测（一）：两条腿、七条规则、档 A 九型）; the cards do not repeat them. ## 5.11 Single-service price page (pt01) **What you'll do in this section**: build one page for every service buyers have already settled on, all from one template: the own-price sentence in the first screen, followed immediately by the price table. These are the two spots ChatGPT copies most often. When you're done, you'll have a copyable single-service price template; change the service name and you can open another page. Intent: the buyer has already chosen a service and is asking "how much does this place charge". Which leg: **mainly ChatGPT**. This is the page type the book's "own sentence" comes from (evidence: see （→ 通用版 A.3 页型实测（一）：两条腿、七条规则、档 A 九型）). One URL per service, the same template (Intro → Evidence block → Decision → FAQ → CTA; empty fields hide automatically). title: ` Cost in Singapore () | `; H1: a statement, ` Price in Singapore ()`, not a question. ```wireframe id=pt01 Figure: Single-service price page; AI most often copies the own-price sentence and the price table right after it H1 | Statement, with the year | req Byline and date line | Named reviewer + month and year, or "Rate card reviewed " | opt Own-price sentence | 1–3 sentences, 40–130 words, opens with the figure | req,cite,law Price table | Right after the price sentence: service tier + defining condition + price + what's included | req,cite,law Tier criteria H2 | First sentence states what decides the tier; one figure per tier | opt Included vs charged-separately table | Item / Fee / Notes | opt Subsidies and payment H2 | Separate fact paragraph, apart from the price table | opt Alternatives H2 | First sentence gives your own price for both options; the market sentence can go here | opt FAQ | Restate the billing rules as complete sentences | opt Related price pages, cross-links | List | opt ``` **Standard column headers**: keep only one set — `Item / Fee / Notes`, or by tier `Tier / Includes / Price`; every other industry-specific column header goes in the industry edition. **Sample sentences**: " costs S$400 for or S$900 for at , with at S$200." On the strict side, every amount is written as the final tax-inclusive price, marked incl. GST. The market sentence is optional for this type; put it in the alternatives block or the FAQ. On the strict side, rewrite it as a variable sentence carrying no peer prices. The minimum viable version keeps only H1, the own-price sentence and the price table, with each row giving "service tier + defining condition + nett price + what's included". No lower bound is set on length (evidence: see （→ 通用版 A.3 页型实测（一）：两条腿、七条规则、档 A 九型）). **Compliance** - Availability by side: see （→ 通用版 5.1 页型总图：46 种、六个家族、三档证据、三侧开放）. Cell-by-cell wording: see （→ 通用版 5.2 三侧速查（一）：先判身份；价格、促销与赠送、结果数字）, （→ 通用版 5.3 三侧速查（二）：证言与评价、比较、榜单、头衔、外链、FAQ 与图注）. - Specific to this page type: on the strict side, the price table must not carry any column that implies results or a timeframe (such as a results cell in a spec strip). Keep only prices and objective process parameters. **Most common mistake**: the price sits only in an image, a booking widget or a JS toggle, so it cannot be found in the source code. Before handing over, search the source with Ctrl+F for the price figures; if you can't find them, send it back. ## 5.12 Price guide page (1): why it is always the first page (pt02) **What you'll do in this section**: stop arguing about what the first page should be. For every buyer type, the first page is always an all-products price guide page: every item you currently sell goes into one master table, the official public benchmark is handled by side, the H1 carries the year and is revised every quarter, and a Chinese version ships at the same time. When you're done, you'll have a build order that is not up for debate, and three hard specs, none of which can be missing. ```mermaid Figure: why the first page is always this one: price questions trigger AI answers most often, the off-site slot is locked by compliance, so all the effort goes to owned pages flowchart LR t["80%+ of price questions trigger overviews"] -->|most common in AI answers| page["First page = all-products price guide"]:::hl gap["Off-site price slots: mostly laudatory titles"]:::warn -->|strict side: no submissions| self["All effort shifts to owned pages"] dens["Citation density counted by page type"] -->|owned pages absorb it too| self self --> page ``` The off-site slot is not empty because the market lacks such pages. Third-party price comparison pages do exist, with side-by-side price tables and a way to submit, but almost all their titles carry Best / Top, and on the strict side a title like that rules out any submission or paid listing; if you are already listed, all you can do is ask the publisher to remove you. Owned pages can absorb this density also because, locally, the price guide pages that actually get cited were written by practitioners themselves in the first place. The off-site slot is downgraded to an opportunistic slot. Evidence: see （→ 通用版 A.2 选点、写页、站外与复测的证据）. ```split Figure: three specs; miss one and it doesn't count as this page; each has an on-the-spot check before launch Spec || How to check before launch Every item you currently sell goes into the same master table, each row linked to its single-service page || Row count = the number of items on sale in the signed-off price list; state the reason for any missing row The official public benchmark is placed by side; write "Not published" where there is no benchmark || The acceptance reviewer opens each official link row by row; if a figure doesn't match, the page does not launch H1 carries the year, revised every quarter, with "Updated" refreshed at the same time || The page register logs the "next revision date"; if not revised by quarter end, the page has expired ``` Spec one: build the thick single-service pages as usual, but the entry page is this one master table. A missing row can only be a service you do not offer or have stopped offering, and it must never be dropped silently. Spec three: when you revise the page, update `dateModified` at the same time and state on the page which rows changed this time. A page not revised by quarter end is downgraded to an expired asset. Spec two follows the stricter side: on the unregulated side you may add a public-benchmark column to the master table; on the strict side, the official benchmark can only be a section of its own, giving the table number and update date, with none of your own prices anywhere in that section and no "we are within the benchmark"; where switch E (peer comparison banned) is on, write no peer figures at all. How to judge the three sides and the switches: see （→ 通用版 0.2 先判你属于哪一侧）. ### The entry question: ask it by side On the strict side the question is "are you willing to publish a fixed final price and tiered, itemised prices"; on the light side the question is "are you willing to publish a fee range and your billing method". **Don't use "willing to publish a range" as the entry bar for everyone.** Writing a range is itself not compliant on the strict side, so using it as the entry test would wrongly rule out a client who is in fact compliant and simply cannot write prices as a range. ### The Chinese version ships at the same time; it is not an extra By default this page ships in both English and Chinese at the same time: separate URL, separate H1, and hreflang tags pointing to each other. The reason: most Chinese-language price questions are still unclaimed, so this is the page most likely to go from 0 seats to at least one within 90 days; for the open-field share, see （→ 通用版 A.2 选点、写页、站外与复测的证据）; for how to write the Chinese page, see （→ 通用版 5.9 内容形态、FAQ、中文页与外语页）. This page's density gate starts from three fact categories: price, subsidies and insurance, and process duration; for the density gate itself, see （→ 通用版 5.8 结论块、密度闸、出处闸）. ## 5.13 Price guide page (2): blueprint, column headers, sentences (pt02) **What you'll do in this section**: build the price guide page from the blueprint. The conclusion block gives an own sentence plus a market sentence, the first H2 is the main price question with the price grid table right below it, and the official public benchmark goes into its own section, placed by side. When you're done, you'll have a price guide page skeleton you can copy directly, and every figure in the conclusion block can be found in the table. Intent: the buyer hasn't chosen a business yet and is asking about pricing across the whole category. Which leg: **both legs cite it, AI Mode more**; when ChatGPT treats it as an overview, it cites the official benchmark alongside it. Length comes from the table, not the word count; there's no lower bound on the number of figures, and outbound links are not a condition for being cited (evidence: see （→ 通用版 A.3 页型实测（一）：两条腿、七条规则、档 A 九型）). title: ` Cost in Singapore (): | `; H1: ` Cost in Singapore ()` or ` Rates in Singapore ()`, not a question. ```wireframe id=pt02 Figure: price guide page block order; AI most often copies the conclusion block and the price grid table under the first H2 title | Cost in Singapore (): | req H1 | Statement, with the year; the Chinese version includes the country | req Byline and date line | Named reviewer + month and year, or "Updated " | opt Conclusion block | 40–130 words: own sentence + market sentence, one of four formats | req,cite,law Data basis line | One line right above the table: sample, source or check date | opt H2 main price question | How Much Does Cost? | req First sentence | At , costs S$X (incl. GST) | req,law Price grid table | / Cost / Suitable for, one figure per cell | req,cite,law Official public benchmark | Its own section, with table number and update date; no own prices inside the section on the strict side | opt,law H2 What affects the price | First sentence is the market sentence, 3–5 factors each with a numbered H3 | opt,law Subsidies, insurance and tax H2 | Separate fact paragraph; government figures carry the table number and update date | opt What's included and excluded H2 | Table or list, 4–8 items | opt FAQ | 5–12 questions, 40–60 words each | opt Related price pages, cross-links + CTA | List, only one CTA | opt ``` The conclusion block can be one of four: a labelled text box, a full price list, a KPI stat card, or a price card. For the three fixed points on the Chinese version, see （→ 通用版 5.9 内容形态、FAQ、中文页与外语页）; the blueprint only notes "includes the country" on the H1 row and does not draw a separate diagram. **Standard column headers**: keep only one set — ` / Cost / Suitable for`, one figure per cell, with the unit and basis. "Market range and own price in the same table" is for the unregulated side only. Write subsidy worked examples separately from the price table, and state that a subsidy is a way of paying, not a price cut. **Sample sentences**: own sentence: "At , costs S$X (incl. GST) for "; market sentence: " generally ranges from $X to $Y, depending on A, B, C". The unregulated side writes this pattern as it is; the light side (switch E is not on) writes it only in a separate fact section with its source, never in the same table as your own price and never next to it (Conservative line (not statute text); see 5.2). On the strict side, and where switch E (peer comparison banned) is on, replace the market sentence with an official-basis sentence: same position, word count and sentence pattern as the market sentence, but describing a regulation or a procedure instead; for how to write it, see （→ 通用版 5.6 两条腿，两种句子）. **Compliance** - Availability by side: see （→ 通用版 5.1 页型总图：46 种、六个家族、三档证据、三侧开放）. Cell-by-cell wording: see （→ 通用版 5.2 三侧速查（一）：先判身份；价格、促销与赠送、结果数字）, （→ 通用版 5.3 三侧速查（二）：证言与评价、比较、榜单、头衔、外链、FAQ 与图注）. - Specific to this page type: besides the conclusion block, the first sentence of the "What affects the price" H2 also counts as a market-sentence slot. On the strict side, and where switch E (peer comparison banned) is on, it too must become an official-basis sentence, or a variable sentence with no peer prices. **Most common mistake**: the first screen only sets the scene without giving a figure, or the first-screen summary doesn't match the table's numbers. The fix: take every figure in the conclusion block from the table, and check them one by one before handing over; when the top card and the table disagree, AI uses the table's figures (evidence: see （→ 通用版 A.3 页型实测（一）：两条腿、七条规则、档 A 九型）). ## 5.14 Official pricing page (pt03) **What you'll do in this section**: for a product with a public price list, fill the /pricing page completely. The price cards sit right under the H1, and none of the four elements is missing: the price unit, the annual and the monthly price, the free-tier cap, and the minimum number of seats. When you're done, you'll have a price source that comparison pages, list pages and third-party pages all link back to. Intent: the buyer asks "how much is per month"; this type also picks up "best " and "A vs B", because when ChatGPT compares products it comes back to this page to check the price. Which leg: **mainly ChatGPT** (evidence: see （→ 通用版 A.3 页型实测（一）：两条腿、七条规则、档 A 九型）). Every product with a public price list must have this page, filled in completely; comparison pages, list pages and third-party pages all link their prices back to it. title: ` Pricing Plans | Starting from $X//month`, no year: this is the exception to rule 4 of （→ 通用版 5.7 所有页型都成立的七条）; H1: ` Pricing` or `Plans for `, not a question. ```wireframe id=pt03 Figure: official pricing page; AI most often copies the price cards and the tax-and-currency footnote H1 | Noun phrase, no year | req Billing toggle | Annual/monthly, annual by default, both prices in the HTML | req Row of price cards | Right under the H1: tier name + price + unit + who it's for + key capability | req,cite Free tier and minimum condition | Free tier states the cap, paid tiers state the minimum seat count | req Tax-and-currency footnote | Prices exclude tax + the billing country decides the currency | req,cite Enterprise card | Fill in the four no-public-price sentences under the button, not measured | opt Add-on table | Add-on / Price / Unit | opt Feature comparison table | Feature / / | opt Market sentence | This product's own range and variables, no competitor prices | opt FAQ | Write the billing rules as complete sentences | opt ``` If any one of the four price-card elements is missing, ChatGPT goes to a third-party page to fill the gap. All four must be in the first-screen HTML, not waiting for a click or JS to appear. There is no length requirement, and no byline is needed; add one extra line, `Prices last checked ` (not measured). For how to write the Enterprise card's four sentences, see （→ 通用版 5.10 页型共用件：数字六项检查、无公开价、计价单位、日期与 schema）. **Standard column headers**: add-on table `Add-on / Price / Unit`; feature comparison table `Feature / $X /month / $Y /month`. **Sample sentences**: keep only one own sentence: "Basic costs $X per seat per month, billed annually; paid plans start from N seats, and prices do not include tax." The market sentence gives only this product's own price range and variables, never a competitor's price (sample, not measured): "Most small teams on pay $X–$Y per seat per month billed annually, depending on plan, number of seats and add-on usage." **Compliance** - Availability by side: see （→ 通用版 5.1 页型总图：46 种、六个家族、三档证据、三侧开放）. Cell-by-cell wording: see （→ 通用版 5.2 三侧速查（一）：先判身份；价格、促销与赠送、结果数字）, （→ 通用版 5.3 三侧速查（二）：证言与评价、比较、榜单、头衔、外链、FAQ 与图注）. - Specific to this page type: none. Availability and wording are all in 5.1 and 5.2: on the strict side, and with switch E (peer comparison banned) on, don't build plan cards or a recommended ranking; build （→ 通用版 5.11 单项目价格页（pt01）） instead; where switch B (legally required fields) is on, build an all-in price page. **Most common mistake**: only a slogan and "Contact sales" sit under the H1, and the price cards need a click or JS to appear. A strong domain's empty-shell pricing page can still get cited. That is a strong-domain exemption, and a weak domain cannot rely on it. ## 5.15 Comparison page (pt04) **What you'll do in this section**: when the buyer is asking "A 还是 B" ("A or B"), organise the comparison page by scenario. The verdict goes into the conclusion block and the first sentence of the first H2, the comparison table sits in the first 25%, each criterion H2's heading names the winner, and you write a separate section for scenarios where "the other one fits better"; on regulated sides, compare only options, never organisations. When you're done, you'll have a comparison page with the verdict up front and a structure organised by scenario, not candidate by candidate. Intent: "A vs B which is better", "A or B for ", "A 还是 B" (A or B). Which leg: **mainly AI Mode**; for the citation ratio between the two legs, see （→ 通用版 5.1 页型总图：46 种、六个家族、三档证据、三侧开放）. When ChatGPT meets a "vs" question, it mostly goes to both sides' official pricing pages, spec pages and help centres and builds its own table, so every figure about your own product on the comparison page must come from the same source and carry the same value as your official pricing page and product page (the same-source point is not measured). Two cases where you should not use this type: on a regulated side, wanting to compare organisations (not allowed; compare options only); or comparing only your own models (that is a category list page). title: ` vs (): Which Is Better for ?`; H1 defaults to a statement, ` vs for ()`; the first H2 is the main question, `Which Is Better for ?`. ```wireframe id=pt04 Figure: comparison page; AI most often copies the conclusion block and the verdict in the first sentence of the first H2 title / H1 | title is a question, H1 is a statement, both carry the year | req Comparison-basis line | Check date + both sides' version or tier | req Conclusion block | 60–120 words, or a by-scenario verdict table | req,cite,law H2 main question | First sentence gives the verdict with its condition, followed right away by the scenario table | req,cite Criteria H2 | Numbered H3s, each first sentence states the threshold that flips the verdict | opt Master comparison table | In the first 25%, one checkable value per cell | req,law Per-criterion H2 × N | The verdict goes into the heading; the first sentence repeats it with a figure | opt Price comparison H2 | Same tier, same currency, same billing cycle | opt Scenarios where the other one fits better H2 | ≥2 specific scenarios, each paired with one public fact about the other side | req Not-a-fit conditions H2 | Skip it if… | opt Data sources and dates | Source + check date for every figure | req FAQ | H3 questions, first sentence answers directly | opt Related comparison pages, cross-links | 3–5 entries | opt ``` On pages where the verdict sits in the middle, only that middle sentence gets copied (evidence: see （→ 通用版 A.3 页型实测（一）：两条腿、七条规则、档 A 九型）), so organise the blocks by scenario, not candidate by candidate. The four formats each have their place: "[Competitor] Alternative" and "You vs [Competitor]" are limited to the unregulated side; the former keeps the "who should switch / who should not" and migration section; "Alternatives" in the plural is really a list, and belongs under （→ 通用版 5.16 榜单页（pt05））; naming yourself inside an "A vs B" page is allowed on the light side, but must come after the comparison and verdict, while regulated sides compare only options and never bring in "us". For comparisons of physical goods, AI Mode prefers to cite video, so make the video first and the web page its companion. ```split Figure: five criteria for not disparaging, which even unregulated-side comparison pages must pass Banned: write this and it's disparaging || Rewrite: put this in the same spot instead Writing only your own strengths, never the other side's || Make "scenarios where the other one fits better" its own H2, with ≥2 specific scenarios Writing the other side's figures with no source at all || Every cell carries a public source and check date; where there's none, write "not published" "better / leading / more powerful" with no metric behind it || Every such adjective must be followed immediately by a checkable metric Not making both sides' basis clear || Write the version and date in the table header Leaving the other side's price up after it's out of date || If an expired price isn't refreshed, delete the cell; label prices as of ``` Self-check for criterion 1: delete the "the other one fits better" section. Does the verdict still hold? If it does, you haven't told the truth. Criterion 2: for any cell with no public value, write "Not published as of "; estimates never go into the table. Criteria 3–5 are not measured; recheck criterion 5 every 30 days. **Standard column headers**: scenario table `If your situation is… / Choose / Deciding fact / Source`; keep only one set for the master comparison table, as a cross-industry example: `Criterion / / / Better for`. One value per cell, with a unit; the same row uses the same tier, the same date, the same currency, the same test conditions; write one line for the data basis above the table. **Sample sentences** (figures are placeholders): own sentence: " Starter costs S$X per user per month billed yearly, with a free tier capped at N users (pricing page, checked )." Market sentence: "For a team of under 10, both tools generally cost $X–$Y per user per month, depending on seat count, annual vs monthly billing, and whether add-ons are needed." The unregulated side writes the market sentence as it is; the light side (switch E (peer comparison banned) is not on) writes it only in a separate fact section with its source (Conservative line (not statute text); see 5.2); the strict side, and anyone with switch E on, do not write it. How to split a comparison page from a concept pillar page: a title containing "哪个更好" ("which is better") belongs to this type; a title containing "有什么区别 / 什么是" ("what's the difference / what is") belongs to the concept pillar page; see （→ 通用版 5.30 档 C 四型：未验证页型）. **Compliance** - Availability by side: see （→ 通用版 5.1 页型总图：46 种、六个家族、三档证据、三侧开放）. Cell-by-cell wording: see （→ 通用版 5.2 三侧速查（一）：先判身份；价格、促销与赠送、结果数字）, （→ 通用版 5.3 三侧速查（二）：证言与评价、比较、榜单、头衔、外链、FAQ 与图注）. - Specific to this page type: the comparison table must not have a peer column or a market-range column (on the strict side, or with switch E (peer comparison banned) on). This is a problem of table structure, not of wording. **Most common mistake**: opening with background and pushing the verdict to the middle, so only that middle sentence gets cited; on a regulated side, turning an option comparison into an organisation comparison: that is a compliance breach, not a layout issue. ## 5.16 List page (pt05) **What you'll do in this section**: first check （→ 通用版 5.1 页型总图：46 种、六个家族、三档证据、三侧开放） and （→ 通用版 5.3 三侧速查（二）：证言与评价、比较、榜单、头衔、外链、FAQ 与图注） to judge whether your side may self-build a list; if it may, write the methodology or the check date into the title or the first screen, write each item's H2 as ": Best for ", make every figure in the at-a-glance table match the item card figure for figure, and disclose it whenever your own business ranks first. When you're done, you'll have a list-page template built to the structure of lists that actually get cited. Intent: "Best for ", " alternatives", "Top N ". Which leg: **mainly AI Mode**; for the citation ratio, see 5.1; everything the ChatGPT leg cites is a third-party list; it did not cite a vendor's own list even once (evidence: see （→ 通用版 A.3 页型实测（一）：两条腿、七条规则、档 A 九型）). A "who's best" question about choosing an organisation does not use this type; see （→ 通用版 0.3 哪些页型对你有限制、改做哪一型）. The condition for getting into ChatGPT: every list ChatGPT cites has the methodology or the check date written into its title or first screen. A list where your own business ranks first can still be cited by AI Mode, but the conflict-of-interest disclosure line must be there. title: ` Best for in Singapore ()`; to get into ChatGPT, also write the methodology or the check date into the title, such as `… (Tested on Models)` or `… Prices Checked `; H1 is a noun phrase that makes a statement, not a question; the first H2 is `Which Is Best for ?`, whose first sentence names the winner with its condition. ```wireframe id=pt05 Figure: list page; AI most often copies the conclusion paragraph that names the picks, and the ": Best for " H2s title / H1 | With the year; the methodology or check date goes into the title or the first screen | req Author, date, check-date line | Write all three | req Conclusion paragraph | Before any H2, names the winner / budget pick directly | req,cite First H2 + at-a-glance table | First sentence repeats the verdict, followed right away by the at-a-glance table, in the first 20% | req Table of contents | Anchor list | opt Methodology | How we tested / How we picked | req Conflict-of-interest disclosure line | Must be written when your own business ranks first | req,law One H2 per item | : Best for | req,cite How to choose | Criteria H2, linking to official standards | opt What's changed since last time | 3–6 items | opt Data sources | Source and date for every price and spec | opt FAQ | Restate the verdict on the winner | opt ``` For an item whose heading gives only a name, the line right below it must be `Best for: …`; the order inside the card is: first-sentence verdict → spec strip → 2–3 pros and 2–3 cons → `Skip it if: …` → the actual price + check date + source. Every figure in the at-a-glance table must match the item card figure for figure, and the price column header carries the check date. **Standard column headers**: keep only one set as a cross-industry example — ` / Best For / / Starting Price / Free Tier or Trial`. **Sample sentences** (figures are placeholders): specific sentence: "The was S$X at on , with a -year local warranty." Market sentence: " suited to generally cost S$X–S$Y, depending on , , and ." **Compliance** - Availability by side: see （→ 通用版 5.1 页型总图：46 种、六个家族、三档证据、三侧开放）. Cell-by-cell wording: see （→ 通用版 5.2 三侧速查（一）：先判身份；价格、促销与赠送、结果数字）, （→ 通用版 5.3 三侧速查（二）：证言与评价、比较、榜单、头衔、外链、FAQ 与图注）. - Specific to this page type: none. Availability and the wording for each side are all in 5.1 and in the list-page figure in 5.3; this card only points there and does not rewrite it side by side. **Most common mistake**: each item has only a name and a description, with no "Best for" verdict sentence and no actual price. On a weak domain, a list written like this gets skipped. ## 5.17 One question, one page (pt06) **What you'll do in this section**: for single-point questions with a definite answer (can I, should I, what's the cap, how long does it take, who's eligible), give each question its own URL, with the H1 using the buyer's original question word for word, the answer paragraph right under the H1, the first sentence giving Yes/No or a figure, and any conditions in the 2nd and 3rd sentences. When you're done, you'll have a batch of short-answer pages that can be checked sentence by sentence against the original government page. ```mermaid Figure: which questions use one question, one page; for those that do, the answer's source decides which leg to focus on flowchart LR q{"Does this question have a definite answer?"} -->|no| skip["Don't use one question, one page"]:::warn q -->|yes| q2{"Is it a remedy or symptom question?"} q2 -->|yes| pt08["Build the remedy page pt08 instead"] q2 -->|no| q3{"Who is the source of the answer?"} q3 -->|official or regulatory| gov["Focus on AI Mode; copy official figures exactly"] q3 -->|only you| own["Own-fact question; may get into ChatGPT"]:::hl ``` Which leg: **government versions get into ChatGPT, commercial versions get into AI Mode** (evidence: see （→ 通用版 A.3 页型实测（一）：两条腿、七条规则、档 A 九型）). For market or regulatory fact questions, the ChatGPT leg is occupied by government pages, so commercial sites should focus on AI Mode; an own-fact question can only get into the ChatGPT leg when you are the sole source (not measured). For questions with no definite answer, ChatGPT cites only government and media; for remedy and symptom questions, write the safety triage first [Safety]; see （→ 通用版 5.19 补救与二次评估页（pt08））. H1: the buyer's original question, word for word, with a local qualifier. This is the only one of the nine types whose H1 is a question; title: `? `, and the year is added only when the answer changes over time and someone maintains it (not measured); the URL slug is the question itself, and one URL answers only one question. ```wireframe id=pt06 Figure: one question, one page, short version; what AI copies is the answer paragraph right under the H1 H1 | The buyer's original question, word for word | req Answer paragraph | First sentence gives Yes/No or a figure, conditions in the 2nd and 3rd sentences, 40–90 words | req,cite,law Boundary sentence | Which part cannot be judged online, and who judges it | opt Source line | Source + check date | req,law Own-fact paragraph | Sub-heading "At ", 1–2 sentences of own facts only | opt Byline and date | One line | opt Related questions | 3–4 items, each linking to its own URL | opt ``` A regulatory fact question with many conditions uses the long version: the answer paragraph comes before the first H2, followed by 4–8 verdict-statement H2s, an optional conditions table, and related questions each linking to their own URL; the short version is 40–300 words, the long version no more than 1,500 words. Most one-question pages need no table; a table is only added when a single answer carries several conditions. **Standard column headers**: for eligibility questions, use the question itself as the row name; for limit questions, use `Item / Limit / Condition / Source`. **Sample sentences** (figures are placeholders): regulatory sentence: "Yes. Under , is allowed up to $X; the remainder is generally paid , and the total depends on ." The figures must match the official page word for word; note the check date. Own sentence: " accepts . For , the final fee is S$X including GST and includes ; is priced after assessment." Write the billing rules as complete sentences: whether tax is included, the bundled price for the first N people, the per-unit price after that, and the local currency. **Compliance** - Availability by side: see （→ 通用版 5.1 页型总图：46 种、六个家族、三档证据、三侧开放）. Cell-by-cell wording: see （→ 通用版 5.2 三侧速查（一）：先判身份；价格、促销与赠送、结果数字）, （→ 通用版 5.3 三侧速查（二）：证言与评价、比较、榜单、头衔、外链、FAQ 与图注）. - Specific to this page type: the answer gives only general information, never an individual diagnosis; questions involving symptoms or complications move to the remedy and second-opinion page (5.19) [Safety]. **Most common mistake**: turning one question, one page into one page with ten questions (really an FAQ collection), or opening the answer with "it depends". ## 5.18 Eligibility and process page (pt07) **What you'll do in this section**: when answering "what conditions are needed, how many visits, how long it takes, how to apply, what to bring", use a task-style statement for the H1, put a five-cell key-value spec strip in the conclusion block, and have the first H2's first sentence give the verdict followed right away by the steps table. When you're done, you'll have a process page that feeds both legs at once: a source for ChatGPT on your own process and duration, and, for AI Mode, a sentence of the form "generally X–Y, depending on A/B/C". Intent: what conditions are needed, how many visits, how long it takes, how to apply, what to bring; if you're only answering a single point, use one question, one page (5.17) instead. Which leg: **mainly ChatGPT**, where government process pages are cited the most; commercial sites' process articles only turn up in AI Mode (evidence: see （→ 通用版 A.3 页型实测（一）：两条腿、七条规则、档 A 九型）). H1: a task-style statement, `How to in Singapore`, or ` in Singapore: Steps, Timeline and Eligibility`, not a question; title = H1 + organisation name, and the year is added only when the rule changes by year and someone maintains it (not measured). ```wireframe id=pt07 Figure: eligibility and process page; AI most often copies the verdict in the first H2's first sentence and the steps table H1 | Task-style statement | req Byline and date line | Named reviewer for personal-safety pages; "Last updated" for others | opt Conclusion block | Five-cell spec strip + own sentence + market sentence | req,law H2 main question | First sentence answers directly, ≤35 words | req,cite Steps table | Step / Result, each row states the duration and the number of visits | req,cite Conditions H2 | Numbered H3s covering every case, or a fact × explanation × when-it-applies table | req Bring-list H2 | 3–8 items | opt Itemised fees H2 | Item or service / Fees | req Parts that cannot be judged online H2 | 40–80 words | opt,law Branches after assessment H2 | Four-column branch table; the price cell states the final price or "confirmed after assessment" | opt Risks and exceptions H2 | General risks only, never written as an individual diagnosis | opt,law Alternatives H2 | Option / When / Time / Price | opt Who makes the final call H2 | Who decides, and at what point, ≤50 words | opt FAQ | H3 questions, 40–60 words each | opt ``` The five cells of the conclusion block are `Total time / Visits or steps / Who it is for / Who it is not for / Who decides`, followed by one own sentence and one market sentence. A government page's conclusion block states only the scope and gives no answer. That is a privilege of government pages; commercial pages should not copy it. When there are many conditions, use the questions as table row names (Who is this for? / How much?), with columns split by group or format, one checkable figure per cell. End the page with a Need help section or 3–4 related-page links. **Standard column headers**: `Item or service / Fees`; `Step / Result`; branch table `If the assessment finds / Next step / Added time / Price (final, incl. GST) or "confirmed after assessment"`. **Sample sentences**: own sentence: "At , takes visits and from to ; whether is needed is decided by after ." Market sentence: " generally takes , depending on , , and ." Measured: asked how long a particular process takes, ChatGPT cited a government page and wrote "typically takes about 4 to 5 months from filing to Final Judgment", and AI Mode wrote almost the same sentence. A range plus a condition is the sentence shape both legs copy. **Compliance** - Availability by side: see （→ 通用版 5.1 页型总图：46 种、六个家族、三档证据、三侧开放）. Cell-by-cell wording: see （→ 通用版 5.2 三侧速查（一）：先判身份；价格、促销与赠送、结果数字）, （→ 通用版 5.3 三侧速查（二）：证言与评价、比较、榜单、头衔、外链、FAQ 与图注）. - Specific to this page type [Safety]: on pages related to personal safety, write who is suitable as "judged by a professional after an assessment", never as a self-check list for the reader. A self-check list amounts to letting readers diagnose themselves. **Most common mistake**: conditions written only as "it depends". ## 5.19 Remedy and second-opinion page (pt08) **What you'll do in this section**: when the buyer is asking "what do I do if it went wrong, who do I complain to, I want a second opinion", first pick a skeleton by risk type. For personal-safety cases use pt08-safety, which leads with what to do now and the red flags; for rights-deadline cases use pt08-deadline, which leads with the urgent deadlines. When you're done, you'll have one of the two skeletons, with its order fixed: not one slot may move. ```mermaid Figure: sort a remedy question by risk type first; each skeleton has a fixed opening that must not move, and the two must never be mixed flowchart LR q{"Which risk type is this remedy question?"} -->|personal safety| s1["pt08-safety: lead with what to do now"]:::hl q -->|rights deadline| d1["pt08-deadline: lead with the urgent deadline"] s1 -->|cause| s2["Appears only after the assessment process"] s1 -->|never| bad["Cause first, reader self-diagnoses"]:::warn d1 -->|before the conflict check clears| d2["Do not take case details"] ``` Intent: what to do when something went wrong or failed; who to complain to when there is a problem; wanting a second opinion. Which leg: both legs cite it, but not the same kind of page (evidence: see （→ 通用版 A.3 页型实测（一）：两条腿、七条规则、档 A 九型）). The subject of the H1 is the buyer's situation, not the organisation; you may state neutrally that earlier treatment took place elsewhere, but these three are banned without exception: assigning blame, disparaging, and any unproven judgement that something failed. ### pt08-safety: personal-safety type This skeleton's order is fixed; not one slot may move [Safety]. The reason: when AI itself answers this kind of question, it also leads with safety, so a page in the same order matches the structure of AI's answer. The sample is only 2 pages; evidence: see （→ 通用版 A.3 页型实测（一）：两条腿、七条规则、档 A 九型）. ```wireframe id=pt08-safety Figure: personal-safety remedy page; not one slot in the order may move; AI copies the first-screen definition sentence and the red-flag table H1 + byline line | H1 states the situation; the line below carries both the author's and the reviewer's names + date | req What to do now | First-screen labelled box: one definition sentence + 3–4 actions | req,cite Red-flag table | If you notice / Do this / How soon, in three levels | req,cite,law Things not to judge yourself | 3–5 items | req Assessment process | Step / What happens / Time + a "please bring" list | req Possible pathways | Conditional branch table; this is the first place on the page where "cause" appears | req,law Itemised fee table | The assessment states the final price; treatment fees state "confirmed in writing after assessment" | req Complaint and regulatory channels | Written only when there's a complaint or harm, one sentence per channel stating what it covers | opt Author bio + References | ≤100 words each | req FAQ | H3 questions, 3–6 items | opt ``` The red-flag table's three levels are: emergency care immediately, same day, this week. ### pt08-deadline: rights-deadline type (not measured) This whole blueprint is not measured: there is no cited page to compare it with, so no block in the figure is marked cite. ```wireframe id=pt08-deadline Figure: rights-deadline remedy page, deadlines first, then pathways (not measured as a whole, so no block is marked cite) H1 + byline | H1 states the situation; the byline gives the practitioner's name + years in practice | req Urgent-deadline table | Deadline / What it applies to / Source | req,law Whether we can take it | Which types, at which stage, we can take on, and which we cannot | req Conflict check | No case details are taken before the check clears | req Documents to bring | Document / Why we need it | opt Scope of the second opinion | Covered / Not covered | req Fee table | Service / Fee / Includes / Excludes | req No interference with your existing representation | We do not advise ending your current engagement, and do not contact your current provider ourselves | req,law FAQ | H3 questions, 3–5 items | opt ``` Every row of the urgent-deadline table cites the original text of the regulation or an official page, with the date retrieved. **Standard column headers**: red-flag table `If you notice / Do this / How soon`; pathway table `If the assessment finds / Possible pathway / Who does it`; deadline table `Deadline / What it applies to / Source`. **Sample sentences**: own sentence: "At , a review with is S$X (incl. GST); any further fee is confirmed in writing after that examination." Market sentence: " is usually managed by , , or , depending on , , and ." Negative conditions in the H2, dual bylines and attached References combine into one general rule: write "when this isn't a fit" into the H2 heading itself, not buried in a paragraph; give both the author's and the reviewer's names; and attach References at the end of the page. **Compliance** - Availability by side: see （→ 通用版 5.1 页型总图：46 种、六个家族、三档证据、三侧开放）. Cell-by-cell wording: see （→ 通用版 5.2 三侧速查（一）：先判身份；价格、促销与赠送、结果数字）, （→ 通用版 5.3 三侧速查（二）：证言与评价、比较、榜单、头衔、外链、FAQ 与图注）. - Specific to this page type [Safety]: no promotional CTA such as a free assessment follows the red-flag section; a professional reviewer's written sign-off is required before launch. That sign-off covers the accuracy of the professional content, not compliance clearance: a written sign-off 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）. **Most common mistake** [Safety]: leading with the cause and letting the reader self-diagnose. That is a safety problem, not a matter of layout preference. A layout mistake: multiple H1s on one page, tagging the summary box, the FAQ and the CTA all as H1. ## 5.20 Entity anchor pages: three subtypes, and the product detail page (pt09) **What you'll do in this section**: sort out which intent each of the three entity anchor subtypes serves. The blueprints for the organisation facts page and the person profile page were already built in chapter 3, so here we only add the product detail page: the first-screen buy box, the feed, structured data and the checkout price must all agree, and specs are written as an HTML table. When you're done, you'll have one more skeleton, for the PDP, plus a rule for deciding which of the three subtypes a question goes to. ```mermaid Figure: each subtype serves one kind of intent; don't use the wrong skeleton flowchart LR q{"Which kind of thing is the buyer asking?"} -->|who's best, how much does this place charge| org["Organisation facts page, General Edition 3.2"] q -->|who is this person, are they credible| person["Person profile page, General Edition 3.3"] q -->|how much is this item, what are its specs| pdp["Product detail page pt09-pdp"]:::hl ``` For the blueprint of the organisation facts page (pt09), see （→ 通用版 3.2 事实表与机构事实页）; for the person profile page (pt09-person), see （→ 通用版 3.3 个人实体页与六处痕迹对齐）. Neither is redrawn here. On pages related to personal safety, the first line under the H1 is always the named reviewer; for how to write this line of titles and credentials, see （→ 通用版 5.3 三侧速查（二）：证言与评价、比较、榜单、头衔、外链、FAQ 与图注）. ### Product detail page (pt09-pdp) Intent: "best X under $Y", "X 多少钱" (how much is X), "where to buy". Which leg: **ChatGPT only**; AI Mode relies on the shopping card generated from the Merchant Center feed, not on the body text (evidence: see （→ 通用版 A.3 页型实测（一）：两条腿、七条规则、档 A 九型）). H1: the product's full name + key specs, on one line. ```wireframe id=pt09-pdp Figure: product detail page; AI copies the first-screen buy box and the HTML spec table H1 | Product's full name + key specs | req First-screen buy box | Actual price / original price / discount / stock / delivery, matching the feed, structured data and checkout price | req,cite,law Selling-point H3 | Heading carries a figure | req Spec table | Specification / Value, an HTML table, never an image | req,cite Warranty | Its own section: length of cover, who's responsible, local service centre | req Returns / trial | Its own section | opt FAQ | H3 questions, 3–5 items | opt ``` The Product structured data fills in only the price and stock visible on the page; the original price and the discount must be real. **Standard column headers**: `Specification / Value`, with rows set by category (such as capacity, battery life, weight, warranty length, local service point). **Sample sentences**: own sentence: "The costs S$X (incl. GST) with , and a -year local warranty." Market sentence: " under S$X in this market generally offer and , depending on ." Fill in the market sentence from current peer data, with the price-check date. The unregulated side writes it as it is; the light side (switch E (peer comparison banned) is not on) writes it only in a separate fact section with its source (Conservative line (not statute text); see 5.2); the strict side, and anyone with switch E on, do not write it. **Compliance** - Availability by side: see （→ 通用版 5.1 页型总图：46 种、六个家族、三档证据、三侧开放）. Cell-by-cell wording: see （→ 通用版 5.2 三侧速查（一）：先判身份；价格、促销与赠送、结果数字）, （→ 通用版 5.3 三侧速查（二）：证言与评价、比较、榜单、头衔、外链、FAQ 与图注）. - Specific to this page type: the price must be the same in all four places (the page, the feed, the structured data and the checkout). Once a price is out of date, take it down; never leave an old price up. **Most common mistake**: the H1 is a slogan, the first screen carries only promotions, and the hard facts (price, specs) are hidden in JS or images, unfindable in the static HTML. A big brand can still get cited on the strength of its authority; a weak domain written the same way gets skipped.