Before publishing, verify the page promise from result to action

Preflight record: name the page’s one honest promise

Start the release gate with a sentence that joins audience state, decision, and completion. Attach the exact title, description, canonical URL, and rendered first screen. PASS means an editor unfamiliar with the draft can infer the same job from those elements. REVIEW means the wording is broader than the evidence but can be narrowed. STOP means the title makes a superlative, completeness, safety, or outcome claim the page cannot substantiate. Google’s people-first content guidance supports an intended-audience and usefulness test; its starter guide supports clear presentation and organization. Neither converts a keyword match into a reader benefit. The audit artifact is a result-to-action promise chain where every link records the relevant copy, evidence owner, status, and correction.

Evidence: Google Search Central; GOV.UK Service Manual

Attach the rendered entry evidence

Do not begin by counting keyword mentions. A page can repeat a phrase perfectly while sending the visitor through incompatible jobs.

Evidence: Google Search Central; Google Search Central

Search-clue checks

Sample page-level queries across a documented period, direct audience questions, and at least one current result context. Group them by decision and list counterexamples. PASS means the selected job is visible in more than one evidence source and fits the site’s qualified audience. REVIEW applies to mixed language that may need a branch or narrower snippet. STOP applies when the draft serves a different stage from the people it deliberately attracts. Record locale, device assumptions, and date for result observations because presentation changes. Do not treat search-engine examples as instructions to copy a competitor’s structure. A cluster of similar words is also not proof of a single motive. The promise chain should explain why this page owns the chosen job and where excluded jobs are routed.

If evidence remains thin, publish a narrower answer or wait; manufactured certainty will make later measurement harder to interpret.

Evidence: Google Search Central; World Wide Web Consortium

First-screen and heading checks

Read the page without its navigation and ask where the bounded answer begins. PASS requires the opening to recognize the situation, give or frame the usable answer, and preview necessary qualifications. REVIEW covers essential context that is present but poorly labelled. STOP covers generic introductions, obstructive commercial blocks, or a promised answer hidden after unrelated material. Inspect the heading outline as a map: labels should describe topic or purpose, reveal the reasoning sequence, and allow readers to locate sections without reading every line. W3C’s guidance on headings and labels grounds this accessibility requirement. Avoid adding the same phrase to each heading; repetition can reduce meaning while giving a superficial appearance of optimization.

Capture the outline and the earliest answer sentence in the promise chain so a reviewer can reproduce the ruling.

Evidence: Google Search Central; GOV.UK Service Manual; World Wide Web Consortium

Evidence and boundary checks

List every material claim that changes a reader’s decision. For each, record source type, relevant date, scope, uncertainty, and whether the conclusion exceeds the cited material. PASS means central claims have proportionate support and limitations appear where needed. REVIEW means a claim is plausible but the available source is indirect or dated. STOP means the page offers legal, financial, health, safety, or guaranteed performance advice beyond its competence and evidence. Distinguish official definitions from independent analysis and illustrative calculations. A long reference list cannot rescue a conclusion none of the sources addresses. Also verify that excluded jobs are stated or linked rather than silently absorbed into the article.

This gate protects intent alignment from becoming a technique for making unsupported content look relevant. Escalate specialized claims to a qualified reviewer when required.

Evidence: Google Search Central; GOV.UK Service Manual

Continuation and commercial-fit checks

Trace the knowledge state immediately before each internal link, signup, download, or product reference. PASS means the action is a foreseeable continuation, its destination is clear, and commercial relevance is disclosed. REVIEW means the destination helps only one branch of a mixed audience and needs better signalling. STOP means urgency, scarcity, or a purchase prompt substitutes for the answer. Ask a representative reader to predict what will happen after activating the control. W3C’s label principles apply because an action should communicate purpose, not merely look prominent. Affiliate disclosure remains necessary but does not make a poorly matched offer useful. The chain should be able to end with ‘no commercial next step’ when the reader job does not warrant one.

Do not use click-through rate alone as the quality ruling; accidental or pressured clicks can move that number in the wrong ethical direction.

Evidence: Google Search Central; Google Search Central

Release, rollback, and review trigger

Release when no STOP finding remains, each REVIEW item has an owner and reason, technical destinations work, and the result-to-action chain can be inspected by another editor. Preserve the prepublication copy and measurement definitions. The next action is to run this gate on one draft and repair the earliest break instead of polishing every section at once. After publication, use task feedback, page-level query trends, and events as separate signals; none proves causation by itself. Search snippets may change, analytics may be incomplete, and small samples can swing. Reopen the gate when the audience state, central evidence, product relationship, or recurring query job changes. A mixed-intent page may pass through explicit branches, while a forced single path may fail. The checklist’s value is a recoverable promise, not an all-green dashboard or a prediction of ranking.