FashionMember’s Editorial and AI Verification Checklist

FashionMember’s goal is to publish useful English-language reporting about fashion, AI, retail, materials, emerging brands, and the Los Angeles Fashion District. Producing frequently does not lower the evidence threshold. It increases the need for a visible, repeatable publish gate.

This checklist is the working standard for an editor-review draft. It does not mean every story is complete or error-free. It requires the editor to show what was checked, separate observation from inference, label synthetic media, and keep a correction path open.

An article cannot pass simply because every box is marked. The underlying evidence must support the public claim.

1. Assignment and reader value

• [ ] The story has a specific reader and useful question.
• [ ] The assignment fits FashionMember’s Fashion × AI, District, business, trend, material, brand, or guide mission.
• [ ] The working title does not promise a result the reporting cannot establish.
• [ ] The editor records why this story should exist beyond summarizing public sources.
• [ ] Conflicts, commercial relationships, affiliate links, gifts, travel support, or sponsored status are declared before reporting.
• [ ] The required verification gate and visual-rights plan are written before drafting.

Google Search Central’s people-first content guidance asks whether content offers original reporting or analysis, provides clear sourcing and authorship, and serves an intended audience rather than search traffic. Search guidance is not an editorial code, but those questions support FashionMember’s decision not to mass-publish commodity summaries.

2. Sources and claim ledger

• [ ] Every factual claim that could change is attached to a source and verification date.
• [ ] Primary sources are used for laws, rules, standards, specifications, schedules, policies, company statements, and measured results wherever available.
• [ ] Independent evidence is added when a company has an interest in the claim.
• [ ] Direct interviews identify the speaker, role, date, method, and permission terms.
• [ ] Anonymous sourcing requires editor approval and a documented reason; the published description protects the source without misleading the reader.
• [ ] Quotes are checked against recording, transcript, notes, or written response.
• [ ] Numbers preserve unit, period, geography, denominator, method, and source.
• [ ] Correlation, forecast, estimate, modeled result, and observed result are labeled differently.
• [ ] Contradictory evidence and material uncertainty are included, not buried in notes.
• [ ] Source links resolve to the page that supports the claim, not a search result or unrelated home page.

For each important claim, the internal ledger records: claim text, source, source type, date accessed, evidence excerpt or location, reporter interpretation, uncertainty, reviewer, and final wording. AI may help organize this ledger but may not be treated as the source.

3. Current-information check

• [ ] Laws, policies, product specifications, platform rules, executives, pricing, dates, schedules, and availability are checked close to publication.
• [ ] The article distinguishes enacted rules from proposals, guidance, standards, and voluntary frameworks.
• [ ] A future effective date is named when relevant.
• [ ] Archived material is not presented as current without explanation.
• [ ] “Latest,” “today,” “now,” “best,” and “first” claims have dated evidence or are removed.
• [ ] Planned stories that depend on interviews, field visits, live tests, or current market samples stay unpublished until that work exists.

A “last verified” date is not decorative. It tells the reader when the editor last rechecked the facts most likely to change.

4. AI-assisted research and writing

• [ ] The editor records which tasks used AI: brainstorming, search planning, transcription, translation, extraction, analysis, drafting, editing, coding, or image work.
• [ ] Every AI-produced factual statement is checked against an inspectable source.
• [ ] Citations are opened and verified; invented, mismatched, or circular citations are removed.
• [ ] Confidential, personal, licensed, embargoed, or source-protected data was not entered without authorization and approved controls.
• [ ] The final article adds reporting, analysis, tests, data, or tools that serve the reader.
• [ ] The byline names the responsible human author or editorial team.
• [ ] A plain-language disclosure explains material AI use when a reader would reasonably want that context.

Google’s guidance on generative AI content emphasizes accuracy, quality, relevance, context, and the risk of scaled content that adds little value. FashionMember applies the higher editorial rule: automation never converts an unverified draft into a publishable fact.

5. Data, code, tests, and recommendations

• [ ] The method, sample, date, environment, and excluded cases are documented.
• [ ] Original data is preserved or its origin and transformation are recorded.
• [ ] Formulas, code, and intermediate outputs are reproducible where practical.
• [ ] Synthetic or fictional inputs are labeled at first mention and in the result.
• [ ] A baseline or control is included when a performance comparison is made.
• [ ] Failures, reviewer disagreements, and negative results are retained.
• [ ] The article does not generalize beyond the sample.
• [ ] A platform preflight is not described as ranking, conversion, customer, or financial performance.
• [ ] Financial, legal, medical, safety, customs, labor, privacy, and other high-stakes guidance receives qualified professional review or clearly remains general education.

NIST’s AI Risk Management Framework Core emphasizes context, benchmarks, testing, documentation, measurement, monitoring, and multidisciplinary perspectives. FashionMember uses those principles to structure AI tests while making clear that the voluntary framework is not a certification.

6. Images, rights, and provenance

• [ ] Every image has a creator, source, rights basis, permitted uses, and expiration where applicable.
• [ ] People shown in original reporting have appropriate consent and context.
• [ ] Product, location, and company captions identify what was actually photographed and when.
• [ ] Retouching does not materially misrepresent a product or reporting subject.
• [ ] AI-generated or materially AI-altered images are clearly labeled in the caption and asset record.
• [ ] Synthetic imagery is never presented as documentary evidence of a real business, person, factory, event, product result, or field visit.
• [ ] Alt text is written for the image’s role in context.
• [ ] Logos, trademarks, artworks, faces, and supplied assets are used only with a documented basis.
• [ ] The editor checks crops, mobile presentation, captions, disclosure, and credit on the final page.

The W3C WAI alt decision tree guides whether an image is informative, functional, complex, redundant, or decorative. The C2PA specification can provide tamper-evident provenance assertions. Neither alt text nor Content Credentials proves the underlying editorial claim; both serve different trust and access needs.

7. Los Angeles Fashion District profiles

• [ ] The business’s legal or public operating name is confirmed directly.
• [ ] Location, website, official social account, category, services, ordering method, and appointment policy are checked with the business.
• [ ] Minimum order, price, lead time, inventory, represented brands, and hours are published only when directly verified and dated.
• [ ] Founder or representative quotes are approved for accuracy, not rewritten into claims they did not make.
• [ ] Environmental, labor, origin, certification, and performance statements have supporting documents and appropriate review.
• [ ] Original or business-supplied images have written usage permission and accurate captions.
• [ ] Sponsored, partner, affiliate, or other commercial relationships are prominently disclosed.
• [ ] The public profile shows its last-verified date, verification method, and correction path.
• [ ] The profile stays in draft if direct verification or image rights are incomplete.

A submission through FashionMember’s “Introduce your business” form creates an editorial lead only. It does not create a listing, guarantee coverage, buy rank, or publish supplied claims automatically.

8. Fairness, harm, and language

• [ ] People and organizations have a fair opportunity to respond to significant criticism.
• [ ] The response deadline, channel, and outcome are recorded.
• [ ] Identity language is relevant, accurate, respectful, and based on self-identification where appropriate.
• [ ] The story avoids stereotypes and does not infer sensitive traits from appearance or data.
• [ ] Claims about workers include worker perspectives or explain why they could not be obtained.
• [ ] Safety, privacy, security, labor, accessibility, and human-rights risks are not reduced to business efficiency.
• [ ] Headlines, thumbnails, notifications, and social copy remain as accurate as the body.

The FTC’s advertising and marketing resource hub reinforces the U.S. requirement that commercial claims be truthful, non-deceptive, and substantiated. Sponsored and affiliate material receives disclosure and review separate from independent editorial selection.

9. Page and accessibility QA

• [ ] Headline, deck, byline, published date, updated date, category, and reading order render correctly.
• [ ] Links are descriptive, functional, and safe.
• [ ] Headings form a logical hierarchy.
• [ ] Images have correct alt treatment, captions, dimensions, and responsive behavior.
• [ ] Tables have headers and remain understandable on small screens or receive an alternative presentation.
• [ ] Color contrast, focus order, keyboard controls, form labels, errors, and touch targets are checked.
• [ ] Structured data matches visible content.
• [ ] The page has no unfinished copy, duplicated paragraph, broken asset, or hidden disclosure.
• [ ] Mobile and desktop previews are reviewed before scheduling.

10. Final publish decision

The responsible editor answers four questions:

1. What is the strongest supported conclusion?
2. What important uncertainty remains?
3. What could harm a person, business, or reader if wrong?
4. What evidence would cause us to correct, update, or remove the story?

Then assign one status:

• Publish: required reporting, rights, review, and page checks pass.
• Publish with explicit limitation: the remaining uncertainty is material but responsibly described and does not make the core publication misleading.
• Hold: evidence, response, rights, review, or technical checks are incomplete.
• Do not publish: the assignment cannot meet the standard or would create disproportionate harm.

The editor’s name, decision date, unresolved limitations, and scheduled recheck are retained internally.

11. Corrections and updates

• [ ] Readers and subjects have a visible corrections channel.
• [ ] Requests are logged with receipt date, claim, evidence, owner, and disposition.
• [ ] Material factual corrections are explained on the page with date and change.
• [ ] Silent edits are limited to immaterial spelling, grammar, or formatting fixes.
• [ ] Updated stories are rechecked beyond the single changed sentence when related facts may also have shifted.
• [ ] Removed content preserves an internal record and receives a public explanation when appropriate.
• [ ] Expired District facts are marked or unpublished until reverified.

Trust is not a permanent badge granted at first publication. It is the ability to show the evidence, describe the method, acknowledge limits, and correct the record.

Sources and verification

• Google Search Central: Creating helpful, reliable, people-first content — official guidance on audience value, originality, authorship, sourcing, and explaining how content was created.
• Google Search Central: Guidance on generative AI content — official accuracy, quality, relevance, disclosure-context, and scaled-content guidance.
• NIST AI RMF Core — official voluntary framework for contextual, measured, monitored AI risk management.
• W3C WAI: Alt Decision Tree — official contextual alternative-text guidance.
• C2PA Specifications — official technical provenance and Content Credentials standard.
• FTC: Advertising and marketing — official truth-in-advertising and substantiation resource hub.