Verifiable Local Architecture · Updated 2026-11-18

How Privacy Works on Receipt Scanner

Privacy claims only matter when you can verify them yourself. Receipt Scanner has no backend API for scanning receipts, no user database, and no server-side OCR queue. Everything runs on your device using static web assets and browser storage.

1. What Happens Inside Your Browser Tab

When you open a photo or PDF invoice on the Scan receipt workspace, all compute executes locally inside the browser sandbox. We serve only static files over Cloudflare CDN, configured with strict security headers so external code cannot silently exfiltrate data.

Workflow StepWhere Execution HappensNetwork Requests Sent
Image resize & contrast stretchLocal Canvas / OffscreenCanvas in browser memoryZero upload requests
Digital PDF text extractionSelf-hosted PDF.js worker on your deviceStatic worker script once per session
Text recognition (OCR)PP-OCRv6 tiny WebAssembly model on CPU threadsZero image uploads (~15 MB static weights once)
Field parsing & tax rulesClient-side TypeScript heuristics & regional packsZero requests
Ledger & receipt photo storageBrowser IndexedDB scoped to receiptscanners.comNo remote syncing

2. Verify Zero Uploads Yourself in 60 Seconds

Rather than taking our word on faith, inspect what the application actually transfers across the network on any desktop browser.

  1. Open Scan receipt and open your browser’s DevTools > Network panel (enable Preserve log so nothing clears between views).
  2. Run a first scan or tap a sample receipt so static assets load and the offline Service Worker caches the ~15 MB WebAssembly OCR runtime.
  3. Switch your device to Airplane Mode or select Offline throttling inside the Network panel.
  4. Drop another receipt image or digital PDF invoice, edit the extracted fields, save it to a book, and export CSV. Notice every step completes normally with zero outbound POST requests.

3. Hard Security Headers & Isolation Boundaries

Every HTML page and static asset on Receipt Scanner ships with strict isolation headers defined in our static host configuration: a hash-locked Content-Security-Policy, Cross-Origin-Opener-Policy: same-origin, and Cross-Origin-Embedder-Policy: require-corp. Those headers allow multithreaded WebAssembly acceleration while blocking unapproved remote frames and third-party scripts.

Technical specifications covering these browser isolation guarantees are maintained publicly in MDN Cross-Origin-Embedder-Policy docs and MDN StorageManager.persist() docs.

4. What Storage Risk Means When There Is No Cloud

Because Receipt Scanner stores nothing on our servers, there is no remote backup copy if your browser clears site data or Safari applies 7-day storage caps in non-installed tabs. Read our detailed guide on keeping local expense records safe, or export periodic JSON/CSV backups inside My expenses.

Software boundary notice: Local processing protects receipt confidentiality, but you are still responsible for exporting backups and must verify extracted totals, dates, and tax fields against the original receipt before reimbursement claims or tax returns. Review our full Privacy Policy and About page for complete legal details.

Frequently asked questions

Do my receipt photos or PDFs ever leave my browser during scanning?

No. Image preprocessing, WASM OCR inference, PDF text extraction, and IndexedDB storage run inside your browser tab on your own hardware. Zero receipt files or extracted totals are posted to external servers.

How can I independently prove zero uploads happen?

Open browser DevTools -> Network tab before selecting a receipt on /scan, or turn on airplane mode after the first load. Once static engine files are cached, scanning completes with zero external OCR API requests.

What static assets load from the network on first visit?

Your browser fetches HTML, CSS, JavaScript bundles, and on your first scan roughly 15 MB of static WASM runtime and PP-OCRv6 model weights from Cloudflare CDN, then caches them offline.

Does Google Analytics see my merchant names or spending totals?

No. Analytics records anonymised page visits and non-identifying feature events only. It never accesses IndexedDB records, receipt images, file names, merchants, dates, or currency amounts.

What happens to my expense records if I clear browser site data?

Because we keep zero server copies, clearing browser storage or losing an unbacked-up device removes local IndexedDB records. Export CSV or JSON backups regularly from My expenses.