AI Usage Disclosure

• AI helps the human concierge work faster. AI does not decide pricing, grading, refund eligibility, or whether to ship your ring.
• You can request a human-only path at any time and it will be honored within one business day.
• We do not use your account or message content to train any third-party model. We do not train any in-house model on customer data.

RingsToday uses large-language-model (LLM) and recommendation tooling, accessed through the Replit AI Integrations proxy, in the following narrow ways:

• Stone matching. Surfacing candidate stones from the catalog that fit a budget, shape preference, or written description. The concierge reviews and curates before anything is presented to you.
• Drafting. Drafting concierge replies, follow-up summaries, and educational copy. A human reviews and edits before send.
• Photo and report description. Producing alt text and accessible descriptions of catalog photos and IGI reports.
• Operational tooling. Internal admin features that summarize a long thread or surface stale collections to the concierge for follow-up.
• Search assistance. Translating a natural-language question (“something close to 1.5 carat, oval, low setting”) into structured catalog filters.

• Setting prices. Catalog and quote prices are set by the human concierge and the atelier.
• Grading diamonds. We rely on independent labs (IGI primarily; selected GIA reports). We do not generate, modify, or interpret grades with AI.
• Approving or denying refunds, returns, exchanges, or warranty claims. The human concierge applies the published Refund Policy and Warranty & Aftercare Policy.
• Deciding whether to ship a finished ring or to release a custom build to the courier.
• Verifying identity for sign-in, payment, or KYC. Identity is verified by Clerk and our payment processor.
• Generating fake stone photos or fake IGI reports. The bucket photos and report numbers shown to you correspond to real graded stones.

When AI assists the concierge, it may receive:

• The current message thread, where you are working with the concierge on a collection.
• The structured collection state (selected diamonds, settings, budget, shape preference).
• Catalog metadata for candidate stones.
• Limited account context (first name, time zone, the public-facing collection nickname).

It does not receive:

• Payment-card numbers or full payment instruments.
• Government-issued identification or KYC documents.
• The precise shipping address, until the human concierge explicitly drafts a shipping notification.

We do not contribute your account, collection, or message content to any third-party model’s training set, and we have configured our LLM provider relationships through the Replit AI Integrations proxy with that constraint in mind. We do not operate an in-house model trained on customer data.

Per-call inference logs and prompts are retained on our side for up to 30 days for debugging and abuse prevention, then deleted or de-identified.

You can ask, at any time and for any reason, that the concierge handle your collection without AI assistance. Email concierge@ringstoday.com with the subject line “Human-only” and we will confirm within one business day. The human-only path may be slightly slower (the concierge writes everything from scratch, searches the catalog manually) but it is otherwise identical.

AI tools occasionally produce wrong answers, made-up facts, or off-tone language. The human concierge is the accountable author of every customer-facing message and recommendation. If you receive a message you believe was sent in error or contains a wrong fact, reply to the concierge or email concierge@ringstoday.com. We will correct the record promptly.

We review this disclosure when our AI usage materially changes (new feature, new vendor, expanded data scope) and at least annually. The current effective date appears at the top of the page.