Data & ownership: the catalog stays yours — we just prepare it for AI

Every analysis is tied to your account. While a scan is anonymous, only someone holding its unique link with a random identifier can open it — it can't be guessed or discovered. Once you claim the analysis to your account, it's yours alone.

• An anonymous scan opens only for whoever holds its unique link with a random identifier — nobody else can find it.
• After it's claimed, an outsider's access attempt returns 404 — we don't even reveal that the analysis exists.
• Omnicado support access is limited and used only to resolve your specific request.

Your production data lives in a managed Postgres database and the app runs on Vercel infrastructure. The exact region and data-residency guarantee we confirm individually — we won't promise more than what's in writing in the contract.

• Production data in a managed Postgres database (Drizzle ORM).
• The application runs on Vercel infrastructure.
• The specific region and data-residency guarantee are confirmed in the contract.

Enrichment and analysis run through large language model APIs (for example OpenAI). We send your catalog to them only to generate suggestions for you — not to build a public model. We commit to not using your catalog to train public models and verify the matching terms with each AI provider.

• We send data to the AI API only for your own analysis and enrichment.
• Commitment: we do not use your catalog to train public models.
• Processing terms with each AI provider are confirmed in a DPA.

Your original production fields — title, description and identifiers — are never rewritten without your approval. Change suggestions go into an approval queue as a “before / after” change, and the originals stay stored separately with version history. The additive AI layer for agents (use-cases, context, intents, comparisons) is generated automatically into a separate AI feed — it doesn't alter your original data.

• Original values stay stored separately with version history — the AI never silently overwrites them.
• You approve suggested changes to production fields in a “before / after” queue — you decide what gets used.
• Bulk “approve safe” covers only filling empty fields and the additive AI layer (use-cases, context, comparisons). Rewriting an existing title or description is always approved manually.
• Sync never writes into your store — it only prepares a new feed version that you or AI agents pull. An action audit trail and rollback are always available.

Reselling someone else's goods and need the feed to keep matching on Heureka or Zboží.cz? We leave the critical pairing fields untouched and optimize only a safe AI layer on top — where it can win you visibility without breaking the matching.

• The AI never generates or changes the critical identifiers EAN, GTIN, MPN and product code — they stay untouched for marketplace matching.
• We optimize a safe AI layer on top: use-cases, context, comparisons and filling empty fields.
• Rewriting the title or description of re-sell products is always left to your approval, so the matching doesn't break.
• On the roadmap: a dedicated re-sell profile that disables rewrites of pairing fields wholesale with a single toggle.

We don't hold your data hostage. You download the feed in machine-readable formats, generate exports for specific AI platforms, and can delete an analysis any time with one click — the delete also removes all related data with it.

• Exports for AI platforms: ChatGPT, Gemini, Perplexity and Claude.
• The feed in JSON and XML (Google Merchant, Heureka) — download it any time.
• Deleting an analysis is complete: one click removes the analysis and all related data (enrichment proposals, product versions, measurements, sync logs, reports) in a single transaction — this also works for anonymous pre-signup scans. The published AI feed is a separate object: you delete it yourself as its signed-in owner, or by proving control with a valid API key of the feed (together with its products and API keys); for older feeds without an owner and without keys we erase on request (info@omnicado.com).

The AI feed is deliberately readable for agents — that's the whole point. But you decide who's allowed in: public access can be locked behind an API key and distribution paused at any time.

• Lock the feed behind an API key — we store only its secure hash and the key can be revoked at any time.
• Pause distribution any time; sync has an action audit log, version history and rollback.
• On the roadmap: durable rate-limiting and detailed access logs for the public feed.