Week 2 · Days 8 to 14. About 6 hours 15 minutes of focused work, free like the rest of the plan.
Week 2 · Days 8 to 14
Five days that make every pagequotable by a machine.
Every day below carries its own timing, tools, steps, worked example, pitfalls and completion test; the graded evidence behind the plan lives on the full plan page.
Day 81 hourEnds with: A two-pile fix list: surface it, or write it.
Run the view-source test on every money page
The view-source test sorts every fix into two piles: content that needs writing and content that merely needs surfacing.
A list of your money pagesHomepage, pricing, top services, the page rivals win.
How to do it
Test the money pages first: homepage, pricing, the top three service or product pages, and the page AI most often cites for rivals.
Search the source for your three most important claims verbatim; a claim that lives in a hero image, a video or a tab that loads on click fails even when the page looks perfect.
Log failures by fix type: text that exists but loads by script is a template fix, text that never existed is a writing task for days 9 to 12.
Do it with your AI assistantAI prompt
Role: you are simulating an AI crawler, which reads only the served HTML of a page and runs no JavaScript.
Context: below is raw view-source HTML from [website].
Task: 1) list the important claims that exist as plain text in this HTML; 2) list what is missing because it only arrives through scripts, embeds or images; 3) rank the missing pieces by how much each matters to a buyer deciding.
Rules: quote the exact missing sentence or fact where you can; ignore styling and boilerplate; if the page holds under 200 words of real text, say so first in one blunt sentence.
Output: three sections in that order, the ranking as a numbered list.
HTML: [paste the page source, or its first few hundred lines].
Worked example · Marnwell Kitchens, an invented brand
Marnwell's showroom page fails completely: every word sits inside a carousel script. The fitting-cost guide passes. The template fix that surfaces the carousel copy as plain HTML rescues six pages in one edit.
What goes wrong
Testing the homepage only, when the money pages are where buyers land.
Judging by the browser inspector, which shows the page after JavaScript has run.
Assuming a sentence you can see is a sentence in the served HTML.
You are done with day 8 when
Every money page has been checked in view-source, not the inspector.
Failures are sorted into template fixes and writing tasks.
The worst offender is named and assigned.
Days 9 to 102 hoursEnds with: Rewritten openings and question-form heading outlines.
Restructure the answers to come first
Answer-first is a discipline of first sentences: each section either answers immediately or it buries the answer for both readers and machines.
Have this open
CMS editing accessThis is an editor's day.
Yesterday's fix listRewrite in the order that list ranked.
How to do it
Rewrite the opening 60 words of each money page to answer the page's own question outright, with the brand named in sentence one.
Turn each section heading into the question a buyer would ask, then make the first sentence beneath it the answer; justification follows, never leads.
Read each section aloud alone, without the page around it: if a section no longer says who it is about, the pronouns have to go.
Do it with your AI assistantAI prompt
Role: you are an editor restructuring a page for AI retrieval, where citations concentrate in the first third of a page.
Context: the page below is from [website]; the buyer is [audience]; the brand is [brand].
Task: 1) produce a new heading outline with one h1 and every section heading phrased as a question that buyer would ask; 2) write a 40 to 60 word direct answer for the page and for each section, with the answer in the first sentence.
Rules: start sections with the name [brand], never a pronoun; keep every true fact and cut nothing that matters; add no claims I did not give you; plain language over marketing tone.
Output: the outline first, then the rewritten openings under each heading.
Page copy: [paste].
Worked example · Marnwell Kitchens, an invented brand
Marnwell's pricing page opened with three paragraphs on craftsmanship. It now opens: a Marnwell fitted kitchen typically costs 9,000 to 24,000 pounds installed, and the rest of the page defends the range. That sentence is the one an assistant lifts.
What goes wrong
Rewriting the whole site in one afternoon, so nothing can be attributed later.
Opening sections with a pronoun, which loses the subject the moment the passage is quoted alone.
Keeping a warm-up paragraph before the answer because it reads nicely to you.
You are done with days 9 to 10 when
Each money page answers its own question inside the first 60 words.
Section headings are phrased as questions buyers ask.
Every section still names its subject when read in isolation.
Days 11 to 121 hour 30 minutesEnds with: Fact-backed pages plus an assigned list of open numbers.
Add facts a machine can check
Checkable beats persuasive: a dated number is quotable by a machine and unfakeable by a competitor, which is the whole moat.
Have this open
Access to real numbersThe job ledger, the CRM, the support queue.
Someone who can answer internal questionsGaps become assigned questions, never invented figures.
How to do it
Sweep each page for the usual suspects: fast, leading, trusted, premium, seamless. Each is a slot where a real number should sit.
Attach a date to every figure you publish; an undated statistic ages into a liability.
Where the number does not exist yet, write the internal question that produces it and assign it, rather than shipping an adjective.
Do it with your AI assistantAI prompt
Role: you are a fact editor hunting vague claims.
Context: below is a page from [website]; the brand is [brand].
Task: find every vague claim (words like fast, leading, powerful, seamless, trusted) and propose a checkable replacement for each: a real number with a date, a named source, a first-hand result, or a quotation from a named person at [brand].
Rules: never invent a figure; where no real number exists yet, write FILL followed by the exact internal question I must answer to get it; keep each replacement to one sentence.
Output: a two-column list, vague claim then replacement, followed by the FILL questions as a separate numbered list.
Page: [paste].
Worked example · Marnwell Kitchens, an invented brand
Trusted by homeowners across Yorkshire becomes 214 kitchens fitted across Yorkshire since 2019, checked against the job ledger. The claim survives extraction, and no rival can copy it.
What goes wrong
Inventing a plausible number to fill a gap, which is the one unrecoverable mistake on this page.
Publishing a figure with no date, so it quietly rots.
Replacing an adjective with a different adjective and calling it done.
You are done with days 11 to 12 when
Every vague claim on the money pages is replaced or marked as a pending question.
Each published figure carries a date.
The outstanding questions have names against them.
Day 131 hourEnds with: Refreshed pages and an honest decision log.
Refresh the pages you want cited, honestly
A refresh is content work with a date attached, and only the content half earns the date.
Have this open
The citation list from days 6 to 7Refresh in citation order, not in preference order.
CMS and sitemap accessThree dates move together or none do.
How to do it
Refresh in citation order: start with the pages AI already cites or nearly cites, not the pages you like most.
Update the substance first, numbers, screenshots, the questions the market now asks, then move the visible date, the schema date and the sitemap together.
Leave stable pages alone and say so in the log; a diary of honest LEAVE decisions is what keeps the site's dates trustworthy.
Do it with your AI assistantAI prompt
Role: you are a content editor planning genuine updates, knowing AI answers favour recently maintained pages.
Context: these pages on [website] matter most for [category] questions: [paste up to ten URLs].
Task: for each page, say what a genuine update would look like this quarter: a statistic to refresh, a section the market has moved past, or a new question the page should answer.
Rules: say plainly which pages need no update at all; no cosmetic suggestions, because I only change a visible date when content genuinely changed; one to three lines per page.
Output: one entry per page in the order given, each starting with the word UPDATE or the word LEAVE.
Worked example · Marnwell Kitchens, an invented brand
Marnwell's 2024 cost guide gets real 2026 prices and a new section on flat-pack versus rigid units, so its date moves. The company history page is accurate and stays dated 2023, deliberately.
What goes wrong
Moving a date without changing the content, which costs the whole site's freshness signal.
Refreshing the pages you enjoy rather than the pages that get cited.
Updating the visible date and forgetting the schema and sitemap dates.
You are done with day 13 when
The priority pages are genuinely updated, not restamped.
Visible date, schema date and sitemap date agree on every page touched.
Every page you deliberately left alone is logged as a decision.
Day 1445 minutesEnds with: A ranked fix list filed with the baseline.
Run a readiness audit and file the fix list
The audit is the week's exam: it converts four days of edits into a ranked list of what still blocks retrieval.
Day 8's page listSame pages, so before and after compare.
How to do it
Audit the same money pages you tested on day 8, so before and after are the same instrument twice.
Fix in impact order and re-run only after the top item is done; a fix list re-audited after every tweak becomes noise.
File the scores beside the week 1 baseline; day 30 reads both.
Do it with your AI assistantAI prompt
Role: you are an AI retrieval system scoring a page.
Context: read [URL or pasted HTML of your most important page]; the brand is [brand]; the market is [category].
Task: score the page 1 to 10 on five axes: text visible in the served HTML, answer-first structure, extractable facts, entity clarity around [brand], heading quality.
Rules: justify every score in one sentence; for every score under 8, give the single highest-impact fix, concrete enough to act on today; end with the three [category] questions this page could plausibly be cited for and the one it answers best now.
Output: a five-row score table, the fixes as a numbered list, then the three questions.
Worked example · Marnwell Kitchens, an invented brand
Marnwell's readiness scores move from the 40s into the 70s everywhere except the showroom page, which waits on a developer. That one DEV item becomes next sprint's first ticket, and the week closes measured.
What goes wrong
Auditing a different set of pages than day 8 tested, so nothing compares.
Re-running the audit after every small edit until the numbers mean nothing.
Starting with the easy fixes rather than the ranked ones.
You are done with day 14 when
The day 8 pages are scored again with the same tool.
Remaining fixes are ranked by impact, with developer items ticketed.
Scores are filed beside the week 1 baseline for the day 30 report.
What this week leaves youYou end week 2 with pages a crawler can quote, answers at the top, and a fix list ranked by impact.
The workbook
The workbook turnsreading into doing.
A printable day-by-day checklist, the 48-prompt starter pack organised by buyer intent, the week 4 outreach template and the baseline tracking sheet. Email one carries the workbook link; seven more follow, paced to the plan’s weeks across thirty days, then the newsletter’s occasional notes.
Questions
Asked about this week, answered straight.
Including the ones where the honest answer costs a sale.
Will rewriting pages answer-first hurt my existing Google rankings?
The controlled evidence points the other way: question-form headings lifted organic traffic about 12% in a split test, and answer-first structure is what Google's own extraction rewards. The honest caveat is that structural edits can move any individual page either way, so rewrite the money pages one at a time and watch Search Console for a fortnight rather than changing the whole site in an afternoon.
Do I need a developer for week 2?
Usually for one class of fix only: content that exists but is injected by scripts, carousels or tab widgets, which is a template change. Everything else in the week, rewriting openings, converting headings to questions, replacing adjectives with numbers, moving dates honestly, is editor work in the CMS.
How do I know a page has passed?
Three checks, all free: your key sentences appear in view-source verbatim; a search for one full sentence in quotes returns your page; and the page's readiness score has moved for the reasons you predicted rather than mysteriously. If a fix moved no score, the fix fixed something that was not the blocker.
Every step of this week, with the graded evidence behind it, sits in the full 30-day plan.
Start here
See where you show up in AI answers today.
Add your site and RankX AI suggests the prompts to track, monitors the keywords that matter and audits your pages, with your first results minutes after you finish setup.