Inspiration
Search is changing shape. More and more buyers no longer scroll ten blue links — they ask an AI and read one synthesized answer. If your brand is not in that answer, you are not "ranked lower." You are absent. We run KHAAN.SI, a live AI answer engine that searches the web in real time and answers with numbered, verifiable sources. While building it, we kept seeing brands that had no idea whether AI answers mention them at all — and no simple way to buy an answer to that question. So we turned our own engine into the storefront: the free preview is the product working, and PayPal makes it purchasable in the same flow.
What it does
- Free preview (the AI half). Enter a brand and a category. The app generates five buyer-style questions ("What is the best AI video generator in 2026?"), samples live answers from the KHAAN.SI search/answer pipeline, and reports, per question: whether the brand was Mentioned or Absent, and which competitors were named instead. It closes with a short AI commentary that only restates the sampled facts. Everything is labeled: single engine, single sample, with a timestamp.
- Checkout (the PayPal half). The full GEO Visibility Audit (founding price $299, list $499) is sold through PayPal Checkout: the order is created and captured server-side with the Orders v2 API, the price is fixed in server code (the client can never name its own price), and the buyer pays with a PayPal sandbox account in the official popup. A confirmation page shows the order id, the PayPal order id, the amount, and delivery terms (full audit by email within 5 business days).
- Honest degradation. If live sampling is unavailable, the preview falls back to a clearly labeled snapshot using a fictional brand — never silent fake "live" data. If PayPal credentials are not configured, the app runs in Demo Mode: the checkout walks the exact same flow against simulated PayPal responses, badged as simulated everywhere it appears.
How we built it
Zero-dependency Node.js service (server.js + lib/): serves the single-page storefront and five JSON endpoints — /api/preview, /api/orders/create, /api/orders/capture, /api/config, /api/health. Orders persist to a JSON store. AI sampling (lib/preview.js): category question templates → POST https://khaan.si/api/ask (our production multi-source search-and-answer pipeline — hosted LLM synthesis, including Cloudflare Workers AI) → normalized string matching for brand/competitor presence. Payments (lib/paypal.js): OAuth2 client-credentials → Orders v2 create/capture, PayPal-Request-Id idempotency, capture verification fallback, amount hard-coded server-side. Frontend uses the PayPal JS SDK Buttons and only ever receives the public client-id from /api/config. Design: single page, black-and-gold, no account, no tracking — preview to paid order in about a minute.
Challenges we ran into
Server-side or it didn't happen. The tempting shortcut is creating orders in the browser; we kept create/capture on the server so price integrity and idempotency are real, not theatrical. AI answers are evidence, handle them like it. A visibility claim is only worth making if it is reproducible: question text, engine, timestamp, and matching rules are all in the output or in lib/preview.js. Design for the demo you can actually run. Credentials, networks, and third-party uptime all fail at the worst moment — labeled fallbacks keep the product honest and keep it demonstrable.
What's next
Swap the sandbox credentials for live ones and this checkout becomes the production order page for the GEO service already priced at khaan.si/geo; widen the audit to multiple engines; add the Audit+Fix and Monitoring SKUs.
Built With
- cloudflare
- developer
- khaan.si
- node.js
- paypal
- platform
