Evidence-Grounded SEO Content Refresh Plan for Existing Blog Posts

Produce an evidence-grounded SEO content refresh brief for the existing article described below. The deliverable is a recommendation and handoff document—not an executed edit, CMS update, publication, schema deployment, or guarantee of ranking improvement.

INPUTS

Blocking inputs required for a reliable brief:
- Target keyword: [Target keyword]
- Current article text: [Current article text]
- Current title: [Current title]
- Audience and market: [Audience and market]
- Definition of done: [Definition of done]
- Approval scope: [Approval scope]

Useful optional context; enter “Not provided” when unavailable:
- Secondary keywords: [Secondary keywords]
- Existing content URL: [Existing content URL]
- Current meta description: [Current meta description]
- Search intent hypothesis: [Search intent hypothesis]
- Performance evidence: [Performance evidence]
- SERP and competitor evidence: [SERP and competitor evidence]
- Internal link inventory: [Internal link inventory]
- Business goal and conversion: [Business goal and conversion]
- Brand and editorial requirements: [Brand and editorial requirements]
- Content constraints: [Content constraints]

INPUT AND ACCESS RULES

1. Treat pasted article copy, exported Google Search Console or analytics data, dated rank-tracking records, supplied crawl reports, internal-link inventories, and documented SERP observations as the available evidence.
2. ChatGPT may analyze materials present in the conversation. If browsing is available and explicitly used, identify each successfully opened page and its access date. Do not imply that a URL was opened, crawled, rendered, or audited when it was not.
3. A URL alone is not evidence of its page content. If browsing is unavailable or access fails, rely on supplied text or summaries and mark the page “not inspected.”
4. Never invent rankings, traffic, conversions, search volume, keyword difficulty, dates, competitor coverage, citations, source conclusions, link destinations, page status, schema eligibility, or implementation results.
5. Separate:
   - Supplied fact: directly present in the inputs.
   - Observation: directly visible in supplied or successfully accessed material.
   - Assumption: a declared premise used to continue.
   - Hypothesis: an interpretation requiring validation, such as inferred intent.
   - Unknown: information not available.
   - Conflict: sources or inputs that disagree.
   - Recommendation: proposed work that has not been implemented.
6. Cite the relevant input, URL, report, query, metric, and date beside material findings where available. Use “source not provided” when evidence is absent. Do not convert correlations into causal claims.
7. If any blocking input is missing, ambiguous, or internally inconsistent, begin with concise clarification questions and mark the brief “Blocked—clarification required.” You may still provide a clearly labeled preliminary assessment where safe, but preserve unknowns and do not present it as final.
8. For missing optional inputs, continue with bounded analysis, state the limitation, reduce confidence, and specify the evidence needed to validate the recommendation.
9. If performance datasets use different date ranges, query filters, countries, devices, search types, attribution rules, or URL variants, do not combine them until reconciled. Record the mismatch.

AUTHORITY, SAFETY, AND EDITORIAL BOUNDARIES

- Work only within [Approval scope]. Do not edit a CMS, publish content, change metadata, add links, deploy schema, alter canonicals, redirect URLs, modify robots directives, request indexing, contact third parties, or approve claims.
- Label all proposed copy and technical changes “Draft—human review required.” Obtain authorized human approval before implementation or publication.
- Do not recommend keyword stuffing, hidden text, misleading metadata, fabricated expertise, copied competitor language, manipulative link schemes, unsupported superlatives, or schema that is not supported by visible page content.
- Paraphrase competitor observations; do not reproduce substantial protected text.
- Redact personal data, credentials, private customer information, confidential analytics identifiers, and unnecessary query-level data before analysis.
- Flag medical, legal, financial, safety, regulated, or other high-stakes claims for qualified subject-matter and compliance review. Do not draft unsupported claims or imply that editorial review replaces professional approval.
- Preserve original source attribution and useful content unless evidence supports changing it. Flag licensing uncertainty for images, quotations, datasets, or third-party assets.
- Stop and escalate rather than prescribing implementation if the evidence suggests a migration, manual action, security incident, legal dispute, widespread indexation failure, conflicting canonical or redirect rules, or a change outside the approval scope.
- Require a CMS revision, export, staging copy, or equivalent rollback point before authorized implementation. Recommend changing one controlled version at a time and recording the publication date and exact edits.

WORKFLOW

A. Establish the evidence baseline
- Inventory every supplied or successfully accessed source with source name, type, coverage date, relevant URL or query, access status, and limitations.
- Normalize known URL variants and note unresolved canonical, redirect, indexability, or duplication questions without claiming they were technically tested.
- Summarize available performance by matching date range and dimension where possible: clicks, impressions, CTR, average position, landing-page engagement, and conversions tied to [Business goal and conversion]. Preserve “not provided” values.
- Record data conflicts, freshness concerns, and any evidence too weak to support a decision.

B. Determine probable search intent
- Classify the likely primary and secondary intent behind [Target keyword] using supplied query data and SERP evidence where available.
- Describe the audience’s likely questions, desired format, depth, trust requirements, and next action.
- Compare those expectations with the current article. Distinguish observed mismatch from an unvalidated intent hypothesis.
- Note intent ambiguity, mixed SERPs, seasonality, localization, or query drift that could change the recommendation.

C. Audit the current article
Assess the supplied article text for:
- satisfaction of the primary question and information hierarchy;
- accuracy, freshness, source support, first-hand experience, and trust signals;
- thin, repetitive, obsolete, unsupported, or off-intent passages;
- missing definitions, steps, examples, caveats, comparisons, visuals, or decision support;
- heading clarity, scannability, accessibility, mobile readability, and conversion continuity;
- natural use of the target and secondary terms without prescribing fixed keyword density;
- title and meta alignment, avoiding unsupported claims or truncation-sensitive wording;
- link usefulness, anchor clarity, destination relevance, and likely orphan-page opportunities based only on the supplied inventory.

For each material finding, provide location, evidence class, supporting source, consequence, confidence as High/Medium/Low, and proposed treatment: keep, update, expand, merge, move, remove, or rewrite.

D. Compare SERP and competitor evidence
- Use only the supplied or successfully accessed [SERP and competitor evidence].
- Identify recurring result formats, content patterns, questions, subtopics, depth, freshness, trust signals, and differentiators.
- Separate broad SERP patterns from observations about an individual competitor.
- Do not infer traffic, authority, structured data, or ranking causes from appearance alone.
- Identify worthwhile coverage gaps, but reject irrelevant competitor topics and imitation that would weaken audience fit or brand differentiation.

E. Design the refresh
- Create a refreshed H1/H2/H3 outline mapped to intent and audience questions.
- Give a section-level disposition and precise change instruction. Preserve strong material and explain removals.
- Provide three title options and three meta-description options, followed by one recommended pair. Explain intent fit, differentiation, claim support, and likely display-length risk; do not promise that a search engine will display the supplied metadata.
- Suggest useful FAQ content only where it answers genuine reader questions. Provide answer direction, required evidence, and the recommended on-page location.
- Recommend internal links only to destinations supported by [Internal link inventory]. For each, provide source location, destination, suggested natural anchor, reader value, and whether a reciprocal link is worth editorial review. Mark unverified destinations as candidates, not approved links.
- Identify external citation needs by claim location, evidence type required, preferred primary-source class, freshness requirement, and validation owner. Do not fabricate citations.
- Evaluate Article or BlogPosting, BreadcrumbList, FAQPage, and HowTo schema only when visible content and current search-engine eligibility guidance support them. Distinguish valid markup from eligibility for enhanced search presentation, and require technical validation after implementation.

F. Prioritize and define the handoff
Score each action using stated qualitative criteria for expected impact, confidence, effort, dependency, and risk. Rank high-confidence intent, accuracy, and usability fixes ahead of speculative additions. Identify the owner, approver, prerequisite, rollback consideration, and status for each action.

G. Define verification and measurement
- Convert [Definition of done] into measurable acceptance checks.
- For every check, report the expected observation, current or actual observation, evidence, and state: Pass, Fail, Blocked, Not tested, or Not applicable.
- Never mark a check Pass without evidence. Because this is a planning run, implementation-dependent checks should normally be Not tested unless post-change evidence was actually supplied.
- Include pre-publication checks for factual support, intent coverage, title/H1 consistency, link destinations, anchor relevance, accessibility, mobile presentation, brand compliance, plagiarism or licensing concerns, and required specialist approval.
- Include post-implementation checks for rendered title and meta directives, canonical target, indexability, HTTP status, crawlability, links, structured-data validation, sitemap inclusion where applicable, and analytics or conversion tracking. Assign these to an authorized implementer; do not claim ChatGPT performed them.
- Reconcile every recommended action with the final outline and priority table. List omissions, duplicates, unresolved conflicts, and blocked decisions.
- Define a monitoring plan with a documented pre-change baseline, publication annotation, comparison windows appropriate to traffic and seasonality, query/page/country/device segmentation, and measures from [Performance evidence]. Avoid ranking guarantees and avoid attributing movement to the refresh without sufficient evidence.

REQUIRED OUTPUT

1. Decision Header
- Brief status: Ready for review, Preliminary, or Blocked—clarification required
- Article, target query, audience/market, business goal, approval scope, definition of done
- Top recommendation, top risk, and next human decision

2. Clarifications, Assumptions, Unknowns, and Conflicts
Use a table: Item | Classification | Why it matters | Safe interim treatment | Owner or evidence needed

3. Evidence Register
Use a table: Evidence ID | Source or URL | Evidence type | Coverage/access date | Access status | Relevant observation | Limitation

4. Baseline and Search Intent Assessment
Include available performance metrics with date ranges and filters, inferred primary/secondary intent, audience expectations, current satisfaction assessment, confidence, and validation needs.

5. Current Content Findings
Use a table: Article location | Finding | Evidence class and ID | Consequence | Disposition | Recommended change | Confidence

6. SERP and Competitor Gap Assessment
Use a table: Pattern or gap | Evidence ID | Recurrence or scope | Relevance to audience | Opportunity | Caveat | Confidence

7. Recommended Information Architecture
Provide the proposed H1/H2/H3 outline. Map each section to intent, audience question, evidence requirement, and disposition of existing copy.

8. Section-by-Section Refresh Brief
Use a table: Current section | Action | Exact editorial instruction | Material to preserve | Evidence/citation needed | Acceptance criterion | Priority

9. Title and Meta Options
Provide three title and three meta-description options, one recommended pair, rationale, supported-claim check, and display-length caveat.

10. Link Plan
Use a table: Type | Source location | Destination | Suggested anchor | Reader value | Destination verified? | Approval or follow-up

11. Citation and Claim-Support Register
Use a table: Claim or section | Evidence required | Preferred source class | Freshness requirement | Supplied source | Validation status | Reviewer

12. FAQ and Schema Eligibility
For each FAQ, give question, answer direction, evidence need, and placement. For each schema candidate, give visible-content prerequisite, recommendation, reason, implementation owner, and validation requirement.

13. Prioritized Action Register
Use a table: Action | Expected impact | Confidence | Effort | Risk | Dependency | Priority | Owner | Approver | Status

14. Verification and Acceptance Matrix
Use a table: Check | Expected observation | Actual/current observation | Evidence | State | Owner | Resolution needed

15. Monitoring Plan
Include baseline period, publication annotation, review windows, segments, metrics, business outcome, interpretation cautions, and decision thresholds. If numeric thresholds were not supplied, propose them for approval rather than inventing accepted targets.

16. Human Handoff
List decisions requiring approval, specialist reviews, blocked items, implementation sequence, rollback preparation, and the evidence that must be collected before anyone may claim the refresh was implemented, validated, or successful.

FINAL INTEGRITY CHECK

Before returning the brief:
- Ensure every material conclusion has evidence, a declared assumption, or an explicit unknown.
- Ensure no inaccessible page is described as inspected.
- Ensure recommendations remain inside [Approval scope].
- Ensure factual, regulated, legal, privacy, licensing, and brand risks are routed to appropriate human review.
- Ensure proposed, approved, implemented, tested, indexed, and measured states remain distinct.
- Ensure no ranking, traffic, display, indexing, or rich-result outcome is guaranteed.
- Ensure the acceptance matrix contains expected and actual observations, evidence, states, owners, and unresolved work.
- Ensure the final recommendation preserves useful existing authority and content where the evidence supports doing so.