Skip to content

Trust, but verify.

Every claim on this page is checkable in your own browser. That is the whole point of the architecture.

Last reviewed: September 16, 2026 · English only for now

Architecture

  • Static Astro site on Cloudflare Pages — no application server, database, accounts or AI model.
  • All parsing runs in your tab: main thread for file decoding, a Web Worker for the extraction engine.
  • No analytics, advertising or tracking scripts; fonts are self-hosted.
  • The only browser storage is the ee-theme display preference.
  • The single network exception is the validator's opt-in mail-server check, which sends domain names only over DNS-over-HTTPS after an explicit click.

How to verify the privacy claim

  1. Open DevTools → Network tab.
  2. Paste a unique canary address such as canary-9f2k@example.com into any tool and run it.
  3. Confirm no request body, URL or header contains your content. The browser may fetch same-origin worker/parser bundles on first use — those carry no user data.

Testing methodology

  • 64+ deterministic unit tests cover extraction, cleaning, validation, dedupe, sorting, analysis, counting, domains, formatting, CSV/TSV conversion, MX logic and report exports.
  • Round-trip tests prove CSV, TSV and JSON exports re-import to the same address set.
  • End-to-end browser tests cover the extractor, validator, counter, file uploads, translations, mobile viewports and the zero-third-party-request guarantee.
  • A post-build link checker verifies all 200+ pages ship with zero broken internal links.
  • Every release runs type checks, unit tests, browser tests and a production build before deploy.

Reference data

Disposable-mailbox and multi-label-suffix lists refresh weekly from the community blocklist and the Public Suffix List via a review pull request. Freshness is a build-time concern, so the zero-network runtime guarantee is unaffected. Counts and refresh dates are published on the Validator and Analyzer pages.

Honest limitations

  • Validation is syntax, domain-shape and curated-list classification — not proof a mailbox exists or accepts mail.
  • Files are capped at 30 MB each, 8 per workspace; validation display shows the first 2,000 rows.
  • Scanned PDFs and images need OCR, which the offline build does not include.
  • Archives (.zip/.rar/.7z), legacy .doc and mailbox stores (.pst/.ost) are rejected with an explicit message, not silently skipped.
  • There is no AI model despite the domain name — results are deterministic and reproducible.

Reporting issues

Security reports: hello@emailextractorai.com with the subject “Security report”. Machine-readable contact lives at /.well-known/security.txt. Please do not attach real customer lists or other sensitive data.

Email Extractor is built and maintained by Srikanth M.

Read the privacy modelSelf-hosting & API plans