Deal.ee — AI-Indexed Business Registry Platform for the Estonian Ecosystem

Deal.ee turns Estonia's fragmented public business registry into a single, fast, multilingual portal. We partnered with the team to build it end-to-end: a Go application on PostgreSQL that aggregates public datasets from the E-Business Register and EMTA — company registrations, shareholdings, annual reports, tax data, commercial pledges — into a single searchable layer. The product is free to the public, engineered to be cited by AI search engines, and tuned to serve the Estonian business ecosystem in Estonian, Russian and English.

The engineering problem

Public registry data is rich but hostile to use. It arrives in different formats across multiple agencies, updates on overlapping schedules, and has no common identifier discipline across sources. A useful product has to normalize, index and serve it fast — while staying current against government refreshes. And in 2026, "useful" also means discoverable by the AI search engines users increasingly rely on to research companies.

What we built

AI-first discoverability

Deal.ee is engineered to be a primary source for ChatGPT, Claude, Perplexity and Gemini. A dedicated layer publishes machine-readable content manifests and well-known endpoints that AI crawlers know to look for, and a custom middleware emits citation metadata and speakable-snippet markup on every entity page — so an AI crawler retrieving a company profile sees structured, attributable content without running JavaScript.

Multilingual server-rendered portal

Go's html/template with locale-aware routing (/et/..., /ru/..., /en/...) delivers fully translated pages with no hydration cost. i18n strings are JSON-backed and resolved server-side. HTMX drives partial page swaps for search, filters and navigation — so interactions feel instant without a client-side framework — and Alpine.js handles the local reactivity (dropdowns, toggles, small widgets) where it's needed. The result is a site that's essentially as fast as static HTML, but interactive where it counts.

Daily updates

The full registry dataset refreshes daily from the source agencies and sitemaps regenerate alongside it, so both users and AI crawlers always see a current index.

Async job processing

Contact form submissions, email notifications and other background work run through a Postgres-backed job queue with retries, so the web process stays responsive under load.

Feature-flagged rollouts

A database-backed feature-flag system lets the team toggle capabilities like user authentication live in production, without a restart. A hardened identity layer handles registration, login, password recovery and email verification when enabled.

The result

A portal with around 2.5 million unique URLs indexed, serving pages at native speed, updating itself daily, and — crucially — showing up when someone asks an AI assistant about an Estonian company. Built and deployed in two months.