Privacy Policy

When you use Gemline, we handle:

• Card photos you upload — front/back images for pre-grading, or the 6-angle capture set for Deep Scan.
• Measurements and scores computed from those photos — centering ratios, corner curves, surface defects you've flagged, and the resulting predicted grades.
• Account info, if you sign in — your email address, and whichever reports you choose to save to your history.
• Defect classifications — when you mark a flagged spot on a Deep Scan as a real defect or a false positive, or manually place a defect on a pre-grade photo, that classification (type, severity, and its rough position on the card) is recorded.
• Centering corrections — on the Centering step, the corner positions the tool auto-detected and the corner positions you end up with after adjusting the guides are both recorded, as coordinates only.

Primarily, to run the tool: measuring your card, computing predicted grades, and — if you sign in — keeping a history you can come back to.

Separately, and only with your consent, we use anonymized defect classifications and centering corrections to improve automatic detection. Right now, spotting scratches, dents, and edge wear relies on a fixed heuristic plus your own manual review, and the inner border starts from a fixed guess rather than a real detection. The goal is to eventually have the tool suggest defects and find borders itself, learned from real corrections people have made — so it needs a dataset of those corrections to learn from.

This training use is opt-in and separate from your account. You'll see a banner asking whether you want to contribute before anything is collected for this purpose, and it applies the same way whether you're signed in or not.

Contributions to the training dataset are stored separately from your personal card history and are not linked back to your account or email. What's stored is limited to the classification or correction itself: for Deep Scan, the shape/size/position signal of the flagged spot and whether you confirmed or rejected it; for pre-grade defects, the type, severity, and normalized position on the card; for centering, the auto-detected and final corner coordinates, normalized to the card's own dimensions rather than actual pixels. We do not currently store a copy of the photo alongside any of these contributions — if that changes, this policy will be updated before it does.

We don't sell data to third parties, and we don't use your photos or classifications for advertising. Card images you upload for grading are stored to run the tool and, if you save a report, to show it back to you — not shared beyond that except when you explicitly create a public share link.

Running Gemline means a few outside services handle data on our behalf:

• Supabase hosts our database, account sign-in, and the storage bucket your card photos are saved to.
• Lovable provides the Google sign-in option and receives basic error-diagnostic data (the error message and which page it happened on — not your photos or measurements) so we can fix bugs.
• TCGdex, an open card database, is called directly from your browser when you search for or identify a card (auto-identify, price guide) or view its market price — this sends your IP address to TCGdex the same way it would to any site you visit directly.
• Google, via Google Tag Manager and Google Analytics — see "Cookies & analytics" below for what this actually sets on your device and how to opt out.

These providers may process data outside your own country, including outside the EU/UK. We haven't put formal transfer safeguards (like Standard Contractual Clauses) in place with each of them — if that matters for your use of Gemline, please reach out before relying on the tool for anything sensitive.

We use Google Tag Manager to load analytics tooling — Google Analytics — on this site. The banner shown on your first visit asks before any analytics cookie is set; declining keeps them off, and nothing analytics-related loads until you accept. This is separate from the training-data consent above — accepting or declining one has no effect on the other.

Once accepted, Google Analytics sets cookies such as _ga and _ga_* to distinguish visitors and sessions. We use this in aggregate, to see how the tool gets used and where to improve it — not to identify you personally. You can withdraw consent at any time by clearing your browser's site data for this domain, which re-prompts the banner on your next visit.

• Using Gemline to pre-grade or measure a card doesn't require an account or any consent beyond this — that part of the tool works the same regardless.
• Contributing to the training dataset is a separate, explicit opt-in. You can decline it and keep using the tool normally.
• You can delete saved cards from your history at any time (with a confirmation step, so it isn't accidental).
• If you're signed in, Account & Privacy lets you download a copy of everything tied to your account, or delete the account entirely — deleting removes your saved cards, Deep Scans, and their photos. It does not remove anonymized training-dataset contributions, since those aren't linked back to your account in the first place (see "What 'anonymized' means here" above).
• For anything that self-service page doesn't cover — including withdrawing a prior training-data consent recorded while signed out — contact us using the details below.

Questions about this policy or a request about your data: reach out via the contact details on the site, or open an issue on the project's repository.