- MIRENA
- Compliance
- AI Output Review Standard
Compliance · Governance standard
Generated analysis and copy must pass factual, contextual, rights, safety, and implementation review.
The standard defines review stages for MIRENA-assisted maps, briefs, audits, drafts, rewrites, links, code, schema candidates, and legal or trust content.
Semantec SEO AI Output Review Standard
The standard states what must happen, who owns the decision, what evidence is required, and what blocks release.
Identify the output and intended use
Record the exact file, page, version, workflow, user, audience, destination, and decision the output is expected to support.
A brainstorming note, internal audit, client deliverable, public article, legal policy, and deployable code do not have the same review burden.
Verify inputs and provenance
Confirm that the input pack was authorised, relevant, current, and sufficient. Remove or quarantine credentials, full card data, unlawful content, and unnecessary sensitive information.
Record which sources, files, URLs, and prior decisions were used.
Factual and source review
Verify names, entities, dates, figures, quotations, URLs, product features, prices, law, policies, providers, claims, and citations against current sources.
Generated sources and quotations are rejected until opened and confirmed.
Context and page-ownership review
Confirm the output fits the site, audience, offer, region, page role, user task, cluster, and internal route. Check for cannibalisation, duplicate intent, off-context expansion, and content that belongs on another page.
Semantic and editorial review
Check the opening answer, entity salience, attribute coverage, passage jobs, heading order, transitions, terminology, qualifiers, examples, tables, and conclusion.
Remove vague filler, invented specificity, repeated patterns, unsupported certainty, and language that hides the real decision.
Rights, privacy, safety, and professional review
Confirm the right to use the inputs and outputs. Check personal data, confidentiality, copyright, trade marks, security, consumer, accessibility, advertising, and regulated or high-stakes implications.
Route the output to a qualified specialist where the subject requires it.
Links, code, and structured data
Open material links and confirm that the destination owns the promised task. Test code in the target environment. Structured-data candidates must match the approved visible page and entity record.
Do not treat source-code validity as proof of rendered, functional, accessible, or eligible implementation.
Outcome and disclosure review
Remove ranking, traffic, feature, citation, lead, sales, revenue, saving, or ROI guarantees. Add proportionate disclosure of AI assistance, commercial relationships, limitations, or illustrative examples where material.
Approval states
| State | Meaning | Next action |
|---|---|---|
| Accepted | Required review passed for the stated purpose and version | Handoff to the named next owner |
| Accepted with qualification | Usable only with visible limits or exclusions | Preserve the qualification |
| Repair required | A bounded defect can be corrected | Repair and rerun affected checks |
| Needs evidence | A material source or owner is missing | Request evidence and hold the claim |
| Specialist review required | Subject exceeds the general editor's authority | Route to qualified owner |
| Blocked | Rights, safety, identity, legality, or implementation cannot support use | Do not release |
| Stale | Material upstream change invalidated the review | Revalidate affected stages |
Audit record and handoff
Store the reviewer, date, output hash or version, findings, changes, unresolved items, approval scope, and next owner.
The handoff must not describe failed, blocked, or stale work as complete.
Controlling guidance and first-party records.
The standard uses current source material but does not convert platform guidance into a ranking guarantee.
Google generative AI guidance
Official accuracy, quality, and relevance guidance.
MIRENA workflow
First-party review and publish gates.
OpenAI usage policies
Official external platform rules.
Continue to the owner of the next control.
Claims, sources, AI review, approval, corrections, conflicts, and rights requests remain separate decisions.
Owned route
AI Usage Policy
Public purpose, reliance, disclosure, and prohibited use.
Open route →
Owned route
Human Approval Policy
Approval authority, object, and release scope.
Open route →
Owned route
Source and Citation Policy
Source verification and provenance.
Open route →
Owned route
Claims and Substantiation
Claim evidence and release states.
Open route →
Owned route
Editorial Standards
Writing, structure, accessibility, links, and maintenance.
Open route →
Owned route
Disclaimer
AI output and outcome limits.
Open route →