Write clear, evidence-backed answers
Write precise, useful sections that help buyers decide and give search systems accurate, verifiable text to process.
Rewrite key page sections so buyers can find the answer, understand the context, and verify the evidence. Treat any change in AI visibility or citations as an outcome to measure, not a guaranteed consequence of formatting.
You will rewrite five priority sections into clear, evidence-backed answers with explicit context and source support.
Clear answers and original evidence help people and search systems understand a page, but no paragraph format, word count, schema type, or wording pattern guarantees retrieval or citation. Google says its AI search features require normal Search eligibility, not special AI rewrites or chunking.
IF YOU SKIP IT / Vague copy and unsupported claims make decisions harder for buyers and give search systems less reliable text to process. Clearer writing helps comprehension, but no paragraph pattern guarantees inclusion or citation.
Clarity is useful; citation mechanics are not promised
Answer-first writing, explicit context, and original evidence serve readers and improve editorial quality. Google does not require special chunking, AI-specific rewrites, AI text files, or special schema for AI Overviews or AI Mode.
- Passage
- A coherent section of page content used here as an editorial unit. Systems may process pages differently; this course does not assume one universal chunk size or citation mechanism.
- Self-contained
- A section that remains accurate when shared alone. This is a clarity and safety test, not a promise of retrieval or citation.
- Entity
- A named company, product, person, place, or concept. Use explicit names when needed for clarity and keep material facts accurate across surfaces.
- Qualifier
- The scope on a claim ('for 5-50 person agencies', 'as of Q2 2026'). Qualifiers make evidence easier to evaluate; they do not make a claim automatically citable.
Each phase feeds the next. Don't skip ahead, the artifact at the end is only trustworthy if every phase ran.
- 01Select
Choose questions tied to revenue, recurring prompts, and factual risk.
- 02Extract
Copy the smallest current passage that attempts to answer each question.
- 03Test
Check directness, scope, mechanism, evidence, limitations, entities, and freshness.
- 04Rewrite
Put the qualified answer first without removing necessary nuance.
- 05Support
Add tables, screenshots, data, examples, definitions, and source links where they reduce ambiguity.
- 06Review
Route product, legal, security, or subject-matter claims to the accountable expert.
- 07Monitor
Track passage changes, search performance, citations, and answer accuracy after publication.
- ✓You now have
Answer passage library
Answer before expanding
Put the direct response where a buyer can find it quickly.
A model never reads your page the way you wrote it. It tears the page into chunks, a heading plus its first paragraphs each, and quotes whichever chunk answers the question best on its own. Your beautiful narrative arc does not survive the tearing. Each block either stands alone or it disappears.
So write like a wire reporter: the answer in the first 40 to 70 words, who and what named, then the depth underneath for humans who keep reading. 'Unlock the power of insights' cannot be quoted by anyone. 'For multi-client agencies, X is a reporting platform that...' can. You are not deleting your depth. You are moving the conclusion to the top where the machine can reach it.
- 01
Replace vague headings with descriptive headings or the buyer question when that reads naturally.
- 02
Answer in the length needed to be complete; include the relevant product, audience, and conditions when ambiguity would otherwise remain.
- 03
Use the following paragraphs for mechanism, proof, alternatives, limitations, and next steps.
- Pick the top 20 pages by Search Console impressions on commercial queries (Performance > Pages, filter to money queries).
- For each page, rewrite each section opener in a two-column doc: current version and clearer version. Use the length needed to answer accurately; lead promptly without deleting nuance.
- Ship the rewrites in the CMS section by section. No redesign, no new URLs, just better first paragraphs.
Each rewritten section answers its heading's question within the first two sentences, and the answer stands alone with the page hidden.
- PRO TIP
Write the direct answer before rereading the old paragraph. Editing old copy pulls you back into its throat-clearing.
- WATCH OUT
Answer-first does not mean depth-last-deleted. Keep the mechanism and proof below the answer; you are reordering, not shortening.
OPTIONAL / AI CO-PILOTRun this lesson with AI, the prompt, sources and expected output+
Choose the business priority, interpret exceptions, protect confidential data, challenge weak evidence, and approve the final decision.
Clean, classify, compare, calculate, and draft rows from supplied evidence. AI may surface patterns; it does not own strategy.
Crawler + custom extraction, plus accuracy gaps from AI answer captures.
| Question | Revenue stage | Prompt frequency | Factual risk | Current page | Priority |
|---|---|---|---|---|---|
| How accurate is automated reporting? | Validate | High | High | /automation/ | P1 |
CLAUDE / CHATGPT PROMPTAnalyze the evidence without outsourcing the decision+
Remove personal or confidential data. Attach the named exports, explain every column, and tell the model when the dataset was collected.
You are assisting a human B2B SaaS SEO and AEO operator with Module 07: Write clear, evidence-backed answers. Current lesson: Answer before expanding Objective: Put the direct response where a buyer can find it quickly. Required artifact: Answer passage library BUSINESS CONTEXT I WILL PROVIDE - Product, category, target market, pricing model, sales motion, and geography - The priority customer segment and the commercial outcome for this 90-day cycle - Passage inventory exported from Crawler + custom extraction - Accuracy gaps exported from AI answer captures - Definitions for any internal fields, stages, scores, and abbreviations TASK 1. Replace vague headings with descriptive headings or the buyer question when that reads naturally. 2. Answer in the length needed to be complete; include the relevant product, audience, and conditions when ambiguity would otherwise remain. 3. Use the following paragraphs for mechanism, proof, alternatives, limitations, and next steps. REQUIRED OUTPUT Return a table using these exact columns: Question | Revenue stage | Prompt frequency | Factual risk | Current page | Priority. For every recommendation, cite the source row, URL, call note, or data point that supports it. Add a confidence column in your analysis: High, Medium, or Low. List missing evidence separately instead of guessing. Finish with a section named HUMAN DECISIONS REQUIRED. RULES - Do not invent search volume, revenue, customer statements, product capabilities, or competitor facts. - Do not treat correlation as causation. - Preserve contradictory evidence and explain why it conflicts. - Do not make the final priority or publishing decision. Prepare the evidence for a human owner. - Use this example only as a format reference, not as evidence: “For multi-client agencies, X automates client reporting by…” is clearer than “Unlock the power of insights,” but clarity alone does not guarantee citation.
Each passage starts with a direct answer. A named human owner must verify this before the lesson is complete.
“For multi-client agencies, X automates client reporting by…” is clearer than “Unlock the power of insights,” but clarity alone does not guarantee citation.
Make claims auditable
Specific evidence helps readers evaluate a claim.
'Based on 42 customer implementations reviewed in Q2 2026' looks like legal boilerplate. It is actually armor. A bounded, dated claim gives the model edges it can preserve and gives the buyer a reason to believe the number, so it travels through answers intact while 'save tons of time' gets dropped at the door.
Audit every claim on the page: source, date, scope. Where a number could be challenged, add the one-line method. This feels pedantic the first afternoon and pays for years, because engines increasingly prefer claims they can check and buyers always did.
Saves tons of time
missing
missing
missing
missing
Teams cut report build time by about 60%
Across 42 customer implementations
Reviewed Q2 2026, stated on the page
Before and after build hours, self-logged
First-party onboarding data, linked
- 01
Separate product facts, customer outcomes, opinions, and external statistics.
- 02
Attach first-party evidence to product claims and link primary sources for external facts.
- 03
Include dates, sample size, and methodology where freshness or scope changes meaning.
- Run the page through its claims: every number gets a source and a date inline ('as of August 2026', 'across 42 implementations').
- Link primary sources for external facts, and add a one-line methodology note wherever a number could be challenged.
- Keep the claims sheet from Module 03 updated; it is the audit trail when someone asks where a number came from.
Every factual claim in your passages carries a source, a date, or a scope, and opinions are no longer dressed as facts.
- PRO TIP
Visible dates beat metadata. 'Verified August 2026' in the text is readable by buyers and models; a hidden timestamp is readable by neither.
- WATCH OUT
An unverifiable superlative ('the fastest', 'the most trusted') is weak evidence and may mislead buyers. Replace it with a measurable claim or cut it.
OPTIONAL / AI CO-PILOTRun this lesson with AI, the prompt, sources and expected output+
Choose the business priority, interpret exceptions, protect confidential data, challenge weak evidence, and approve the final decision.
Clean, classify, compare, calculate, and draft rows from supplied evidence. AI may surface patterns; it does not own strategy.
AI answer captures, plus reviewed claims from Product / policy docs.
| Question | Direct answer | Scope | Mechanism | Evidence | Limit | Score |
|---|---|---|---|---|---|---|
| How accurate is it? | Pass | Partial | Pass | Fail | Fail | 3 / 6 |
CLAUDE / CHATGPT PROMPTAnalyze the evidence without outsourcing the decision+
Remove personal or confidential data. Attach the named exports, explain every column, and tell the model when the dataset was collected.
You are assisting a human B2B SaaS SEO and AEO operator with Module 07: Write clear, evidence-backed answers. Current lesson: Make claims auditable Objective: Specific evidence helps readers evaluate a claim. Required artifact: Answer passage library BUSINESS CONTEXT I WILL PROVIDE - Product, category, target market, pricing model, sales motion, and geography - The priority customer segment and the commercial outcome for this 90-day cycle - Accuracy gaps exported from AI answer captures - Reviewed claims exported from Product / policy docs - Definitions for any internal fields, stages, scores, and abbreviations TASK 1. Separate product facts, customer outcomes, opinions, and external statistics. 2. Attach first-party evidence to product claims and link primary sources for external facts. 3. Include dates, sample size, and methodology where freshness or scope changes meaning. REQUIRED OUTPUT Return a table using these exact columns: Question | Direct answer | Scope | Mechanism | Evidence | Limit | Score. For every recommendation, cite the source row, URL, call note, or data point that supports it. Add a confidence column in your analysis: High, Medium, or Low. List missing evidence separately instead of guessing. Finish with a section named HUMAN DECISIONS REQUIRED. RULES - Do not invent search volume, revenue, customer statements, product capabilities, or competitor facts. - Do not treat correlation as causation. - Preserve contradictory evidence and explain why it conflicts. - Do not make the final priority or publishing decision. Prepare the evidence for a human owner. - Use this example only as a format reference, not as evidence: “Based on 42 customer implementations reviewed in Q2 2026…” defines the evidence boundary; it does not make the claim automatically citable.
Entities and audience are explicit. A named human owner must verify this before the lesson is complete.
“Based on 42 customer implementations reviewed in Q2 2026…” defines the evidence boundary; it does not make the claim automatically citable.
Clarify entities and relationships
Use explicit names when context could otherwise be ambiguous.
Lift any paragraph off your site and read it alone. Who is 'we'? The model lifting that chunk does not know either, and your best writing gets attributed to nobody. Pronouns do not survive the tearing, and neither does a category that changes costume between your site, your G2 profile, and your LinkedIn page.
Name yourself in the sentences that matter: product and category, once per section, the same wording everywhere. It reads slightly stiff to a human editor and perfectly to a retrieval system, and the retrieval system is currently deciding whether you exist. When site and profiles disagree, the model averages you into fog, which is why the battlecard in Module 08 exists.
- 01
Name the company, product, category, audience, and integrations where a reader needs that context; use pronouns naturally after the subject is clear.
- 02
Keep material facts consistent across key profiles without forcing identical marketing copy on every platform.
- 03
Use structured data only when it accurately reflects visible page content; no special AI schema is required.
- Review 'we', 'our', and 'it' at section openings. Replace only the ambiguous uses with the product or company name; keep natural pronouns where the subject is clear.
- Cross-check category, audience, capabilities, integrations, pricing state, and company facts against G2, Capterra, LinkedIn, and partner profiles. Correct factual conflicts without demanding identical prose.
- Add or fix structured data only where it mirrors visible text and qualifies for the relevant feature. Google requires no special AI schema or AI text file for AI Overviews or AI Mode.
A reader can identify the subject and evidence without guessing, and material facts agree across profiles even when the wording differs.
- PRO TIP
Introduce the subject when context needs it, then write naturally. Repeating the product and category at every section is unnecessary.
- WATCH OUT
Conflicting material facts confuse buyers and weaken your own QA. The battlecard in Module 08.5 keeps facts aligned without forcing every surface to use identical copy.
OPTIONAL / AI CO-PILOTRun this lesson with AI, the prompt, sources and expected output+
Choose the business priority, interpret exceptions, protect confidential data, challenge weak evidence, and approve the final decision.
Clean, classify, compare, calculate, and draft rows from supplied evidence. AI may surface patterns; it does not own strategy.
Product / policy docs, plus search validation from GSC.
| Question | Answer first | Mechanism | Evidence | Qualifier | Source | Reviewer |
|---|---|---|---|---|---|---|
| How accurate is it? | Automated reports match connected-source data. | Scheduled API sync | QA logs | Subject to source freshness | Product docs | Product lead |
CLAUDE / CHATGPT PROMPTAnalyze the evidence without outsourcing the decision+
Remove personal or confidential data. Attach the named exports, explain every column, and tell the model when the dataset was collected.
You are assisting a human B2B SaaS SEO and AEO operator with Module 07: Write clear, evidence-backed answers. Current lesson: Clarify entities and relationships Objective: Use explicit names when context could otherwise be ambiguous. Required artifact: Answer passage library BUSINESS CONTEXT I WILL PROVIDE - Product, category, target market, pricing model, sales motion, and geography - The priority customer segment and the commercial outcome for this 90-day cycle - Reviewed claims exported from Product / policy docs - Search validation exported from GSC - Definitions for any internal fields, stages, scores, and abbreviations TASK 1. Name the company, product, category, audience, and integrations where a reader needs that context; use pronouns naturally after the subject is clear. 2. Keep material facts consistent across key profiles without forcing identical marketing copy on every platform. 3. Use structured data only when it accurately reflects visible page content; no special AI schema is required. REQUIRED OUTPUT Return a table using these exact columns: Question | Answer first | Mechanism | Evidence | Qualifier | Source | Reviewer. For every recommendation, cite the source row, URL, call note, or data point that supports it. Add a confidence column in your analysis: High, Medium, or Low. List missing evidence separately instead of guessing. Finish with a section named HUMAN DECISIONS REQUIRED. RULES - Do not invent search volume, revenue, customer statements, product capabilities, or competitor facts. - Do not treat correlation as causation. - Preserve contradictory evidence and explain why it conflicts. - Do not make the final priority or publishing decision. Prepare the evidence for a human owner. - Use this example only as a format reference, not as evidence: Introduce “Acme Analytics, a B2B marketing attribution platform,” once, then use natural pronouns while the subject remains clear.
Claims link to suitable evidence. A named human owner must verify this before the lesson is complete.
Introduce “Acme Analytics, a B2B marketing attribution platform,” once, then use natural pronouns while the subject remains clear.
Format for comprehension and verification
Structure should help the buyer scan, compare, and verify.
Tables get lifted whole into answers, framing and all, which makes a good table the highest-leverage block on a commercial page. 'Good' has a definition: columns that name real criteria, like 'Native HubSpot field mapping', not a grid of green checkmarks against vague features. Models skip decoration and so do buyers.
Then run the final exam on every section: hand a colleague one block with the rest of the page hidden and ask what company it is about, what it costs, and who it is for. Whatever they cannot answer, a model cannot quote. Fix it before one tries.
- 01
Use descriptive headings, readable paragraphs, real comparison tables, ordered processes, and labeled definitions where they improve the page.
- 02
Give tables explicit criteria and avoid checkmarks without meaning.
- 03
Read each section for accuracy in context and in isolation; revise anything misleading, then test visibility as an observed outcome rather than assuming the format caused it.
- Add one real comparison or spec table per money page, with column headers that name concrete criteria ('Native HubSpot field mapping', not 'Integrations').
- Break walls of text into readable paragraphs with descriptive headings based on meaning, not a fixed line count.
- Run the read-alone test: hand a colleague one section with the rest hidden and ask what company it is about, what it costs, and who it is for. Fix whatever they cannot answer.
Your commercial pages each contain one real table, critical facts appear everywhere they are relevant, and every section passes the read-alone test.
- PRO TIP
A fair table with concrete criteria helps buyers compare and creates structured visible text. Test whether engines use it; do not assume they will.
- WATCH OUT
A grid of green checkmarks against vague features gives buyers little evidence. Use specific cells or omit the table.
OPTIONAL / AI CO-PILOTRun this lesson with AI, the prompt, sources and expected output+
Choose the business priority, interpret exceptions, protect confidential data, challenge weak evidence, and approve the final decision.
Clean, classify, compare, calculate, and draft rows from supplied evidence. AI may surface patterns; it does not own strategy.
GSC, plus passage inventory from Crawler + custom extraction.
| Page | Passage changed | Published | Indexed | Cited before | Cited after | Accuracy result |
|---|---|---|---|---|---|---|
| /reporting-qa/ | Accuracy definition | 2026-07-12 | Yes | No | Yes | Correct |
CLAUDE / CHATGPT PROMPTAnalyze the evidence without outsourcing the decision+
Remove personal or confidential data. Attach the named exports, explain every column, and tell the model when the dataset was collected.
You are assisting a human B2B SaaS SEO and AEO operator with Module 07: Write clear, evidence-backed answers. Current lesson: Format for comprehension and verification Objective: Structure should help the buyer scan, compare, and verify. Required artifact: Answer passage library BUSINESS CONTEXT I WILL PROVIDE - Product, category, target market, pricing model, sales motion, and geography - The priority customer segment and the commercial outcome for this 90-day cycle - Search validation exported from GSC - Passage inventory exported from Crawler + custom extraction - Definitions for any internal fields, stages, scores, and abbreviations TASK 1. Use descriptive headings, readable paragraphs, real comparison tables, ordered processes, and labeled definitions where they improve the page. 2. Give tables explicit criteria and avoid checkmarks without meaning. 3. Read each section for accuracy in context and in isolation; revise anything misleading, then test visibility as an observed outcome rather than assuming the format caused it. REQUIRED OUTPUT Return a table using these exact columns: Page | Passage changed | Published | Indexed | Cited before | Cited after | Accuracy result. For every recommendation, cite the source row, URL, call note, or data point that supports it. Add a confidence column in your analysis: High, Medium, or Low. List missing evidence separately instead of guessing. Finish with a section named HUMAN DECISIONS REQUIRED. RULES - Do not invent search volume, revenue, customer statements, product capabilities, or competitor facts. - Do not treat correlation as causation. - Preserve contradictory evidence and explain why it conflicts. - Do not make the final priority or publishing decision. Prepare the evidence for a human owner. - Use this example only as a format reference, not as evidence: A comparison table uses “Native HubSpot field mapping” rather than a generic “Integrations” column because buyers can verify the distinction.
The passage remains accurate when read alone. A named human owner must verify this before the lesson is complete.
A comparison table uses “Native HubSpot field mapping” rather than a generic “Integrations” column because buyers can verify the distinction.
The screen below is a clarity and evidence QA example, not a simulated citation mechanism. The stronger version is easier to understand and verify because it names the subject, scope, date, and evidence. Whether an engine includes or cites it must be tested.
“We help teams unlock the power of their data and transform reporting workflows, saving tons of time every month.”
No entity, no audience, no number, nothing checkable.“Acme Reports is a client-reporting platform for 5–50 person agencies. Across 42 implementations reviewed in Q2 2026, teams cut monthly reporting from 2 days to under 3 hours.”
Subject + audience + dated evidence = easier to evaluate.Does the page answer the buyer accurately? Is Google able to index it? Do repeated engine tests include or cite it? Record the outcome; do not assume it.
The workspace shows the passage pipeline: the priority questions, the quality audit against six criteria, and a before/after rewrite. Study the rewrite row. The "after" version answers the question in the first sentence, names the product, and carries its evidence with it. That is the pattern for all five of yours.
Passage engineering lab
Evaluate whether important page sections provide a direct, qualified, evidenced answer that can survive extraction.
Passage quality by component
A quotable section needs more than concise wording.
Check each item only after the artifact meets the standard. Progress saves on this device; use the academy-home backup to move it elsewhere.