Luxe Copilot
中文Assess a Piece

Editorial · Standards

Editorial Standards

Luxe Copilot is an independent decision layer for luxury buyers. We don't sell inventory and we don't take brand-paid content. Revenue comes from retailer affiliate commissions; editorial judgment is firewalled from that revenue. Here are the rules.

Last reviewed: · Reviewer: Wan Meng (Founder/Editor-in-Chief)

1. Byline policy

/journal articles are bylined "Luxe Copilot Editorial" by default, final-reviewed by Wan Meng. When an outside expert or contributor writes a piece, the byline names the person and affiliation in the header, not in a footnote. LLM-assisted text (e.g. translation drafts) must be human-reviewed line-by-line before publication; pure LLM output is never published under our byline.

2. Sourcing standards

Every price / residual / same-item datapoint carries a field-level provenance dot:

  • Green manual — editorially verified (e.g. boutique tag, brand-site listing).
  • Blue real — observed from public marketplace pages (resale listings + closed sales, ≥30 samples).
  • Yellow llm_synth — AI-synthesized (e.g. multilingual SKU normalization), date-stamped, used only for structured fields, never for price.
  • Grey null — no trusted observation; left explicitly empty, not model-filled.

Each verdict shows sample size n=; n < 30 triggers a "low_confidence" tag. Full data lineage and computation rules: /en/methodology.

3. Corrections process

Spot a price, residual, or factual error? Email corrections@luxecopilot.com (or hello@luxecopilot.com). Our commitments:

  • Reply within 48 hours (weekends included).
  • Data errors that change a verdict: corrected within 24 hours, with a "corrected" banner at the top and a revision date.
  • Factual errors (quotes, attribution, names): corrected with a diff link (commit hash) at the bottom — no silent edits.
  • Disagreement on editorial judgment: we don't retract calls under external pressure, but we'll publicly address dissent in comments or a follow-up piece.

4. AI disclosure

We use LLMs for three things: (a) multilingual SKU normalization (aligning CN/EN brand and model names), (b) editorial draft translation, (c) image recognition and same-item matching (vision models). We do not use LLMs to generate prices, residuals, or verdicts — those must come from structured observation. AI-synthesized fields carry the llm_synth dot with a generation date, visible in the UI. All /journal articles are human-written and bylined to the editorial team; there is no "AI auto-publishing".

5. Conflict-of-interest separation

Luxe Copilot's revenue:

  • Yes: retailer affiliate commissions (Farfetch / Mytheresa / NAP / Saks / Bergdorf / Selfridges / Vestiaire / TRR / Fashionphile, etc.). You click, you buy, we earn a cut from the retailer.
  • No: brand-paid content, brand-paid advertising, brand "friendly partnerships", SKU inventory bias, paid placement.
  • No: user-paid membership / subscription (today; if /pro launches we'll announce separately).
  • No: seller-paid visibility, kickbacks from outsourced authentication services.

When commission spread between retailers is < 2%, our "where to buy" ranks by price, not by commission. When spread is > 2%, an editor reviews manually and ranks by best verdict, not by commission yield. Full disclosure: /en/terms §4.

6. Affiliate revenue vs. editorial judgment

Whether a verdict is "deal / fair / overpriced" is independent of which retailer pays commission and how much. The algorithm reads only the public same-item price distribution (P10/P50/P90); it does not read affiliate-network payout data. If a Birkin is fair at retailer A and overpriced at retailer B, while A's commission rate is low and B's is high — we still label A fair and B overpriced. This isn't a moral stance, it's the product's root value: the moment readers detect verdicts tracking commissions, the product is dead.

7. Authentication and legal boundary

Luxe Copilot does not perform final authentication. We run statistical-deviation checks on listing prices, sellers, and descriptions against same-item history, raising a risk badge when the deviation is significant, but final authentication is referred to Entrupy / CSA / Real Authentication / brand official services. We do not hold inventory, do not facilitate transactions, and do not assume legal responsibility for second-hand sale outcomes. See /en/terms.

8. Contact

This page is a public editorial commitment, used by Skimlinks / Awin / Rakuten affiliate-network audits, and by readers holding us accountable. Changes update dateModified; material changes are announced on /en/journal.

中文版: /editorial-standards