Review the actual final draft
Review the exact version you intend to publish. An earlier approved outline or good research packet does not establish the quality of the final article. Compare the draft against the brief, explicit instructions, verified sources, and the reader’s real question.
Review this draft against its brief, supplied first-party evidence,
and SEO Researcher job [job ID]. Do not rewrite or publish yet.
Identify the exact passage for every issue. Classify it as a factual
error, unsupported claim, intent mismatch, unhelpful repetition,
missing useful evidence, or presentation problem. Give the correction
and supporting source. Separate publication blockers from optional
improvements. Do not award a reassuring score over a material defect.Require defensible answers
- Can a reader make the intended decision after reading?
- Does each material claim have direct, current support?
- Are examples, expertise, and test results real?
- Are limitations and important exceptions explained?
- Does the page contribute something useful beyond existing results?
- Are entity names, units, dates, comparison criteria, and product facts consistent?
- Does the final version honor the actual editorial instructions?
If the reviewer cannot access a cited source, classify that claim as unverified. A source URL alone is not supporting evidence. Distinguish small style improvements from blockers; do not bury a consequential false claim under a long list of polish suggestions.
Before publication
Check the final rendered page: title, heading order, links, readability on mobile, accessible images, indexability, canonical URL, and any applicable structured data. Ensure the page and metadata make the same claims. Publication remains a separate action; these skills do not grant publishing permission.
After publication
Record the page version, date, intended queries, audience, and business goal. Use your Search Console, analytics, and any separate AI-monitoring system to compare outcomes. Google reports traffic from AI features within its overall Web search reporting, so do not label all those clicks as traditional-only or claim an AI-only split from that report. Google’s reporting explanation.
For AI observations, retain the exact prompt and answer, platform/model when available, date, and sources. Repeat comparable measurements and report variability. Keep traffic, citations, sentiment, and conversions separate. A single screenshot is an observation, not a durable trend or causal test.