Frequently asked questions

Common questions about Grail Atlas, the scoring, the trust model, and what we do and don't do. Want one added? Contact channel.

Is Grail Atlas a marketplace?

No. Grail Atlas is an information-first reference. We don't hold inventory, we don't broker sales, we don't take payment from sellers. We score listings that exist on other marketplaces and link out.

How does the deal-tier scoring work?

Every listing's landed cost (sticker plus duties, brokerage, VAT, verification fees) is compared to a fair-value band built from comparable verified sales. Below the band median by 10%+ is a great deal; 3-10% below is good; ±7% around median is fair; above is overpriced. See /how-it-works for the full pipeline.

Why don't you call out scams?

Risk is graded (clear, caution, elevated, high) and built from factual, non-accusatory signals. We never call any specific seller a fraud or scammer — that's both a legal exposure for us and a harassment vector against people who haven't been proven guilty of anything. See /guides/counterfeit-awareness for the buyer-side discipline.

How does Grail Atlas make money?

Currently: nothing. The project is self-funded. Future monetization channels (affiliate links on outbound platform clicks, membership for advanced features, verified-listing fees) are documented at /legal/disclosure. The trust composite, valuation engine, and risk read are architecturally separated from any monetization channel.

Where does the comparable-sales data come from?

Currently: a curated seed corpus. Production ingestion (eBay Browse API + selected auction archives) is the next major slice; the spec lives in docs/specs/2026-05-20-ebay-ingest-worker.md.

Why doesn't my reference appear on the site?

The catalog grows as we hand-author editorial bodies for each reference. Send a suggestion via the contact channel in the AUP; references with good market depth and at least 10 comparable sales get prioritized.

Are the prices on the site current?

Reference market-value numbers come from comparable-sales data computed at render time. The recency window is 12 months. The momentum readout shows direction over the most recent 90 days vs the prior 90 days. We never display a precise dollar quote without a confidence label.

Can I trust the valuations for resale purposes?

The valuations are honest estimates from public sold data, with confidence labels attached. They're a good starting point for understanding a fair price. They are NOT an appraisal — for insurance or resale you should use a certified appraiser.

How do I save searches?

Create an account and visit /account/saved-searches. Searches are row-level-isolated to your account (no other user can see them); email alerts come once Slice 1c ships email delivery.

How do I delete my account?

Contact the AUP channel and we'll erase your account data in line with the Privacy Policy. The score-audit trail is the documented exception (legitimate-interest legal-defense override); see the standalone LIA doc for the rationale.

Do you have an API?

Yes. /api/deals for the deal feed, /api/reference/[id] for per-reference data, /api/brands, /api/categories, /api/movements, /api/search. All rate-limited and edge-cached. JSON.

Why no AI image recognition?

AI photo authentication is on the roadmap (Slice 2). We're explicit that it will never assert authenticity — it will surface inspection signals. AI-extracted editorial content is off the table: legal hosting immunity gets compromised when a hosting platform substantially-modifies user-supplied content via model.