Email Spam Score Checker

We run 21 body-content rules adapted from SpamAssassin's URI/SUBJECT/HTML rule families. The aggregate is converted to a 0-10 user-facing score where 10 is best (mail-tester convention). For a real send-time check including headers, IP reputation, and authentication (SPF/DKIM/DMARC), use mail-tester (sends a test message to a unique address) or run SpamAssassin against your finished MIME message. This tool focuses on the part of deliverability you can fix at design time: the body.

Each triggered rule contributes SpamAssassin-style points. The standard "spam" threshold in SpamAssassin is 5.0 cumulative points - we mirror that. A user-facing score of 5 or below means your body crossed the spam threshold from rules alone. 8-10 means clean body. 6-7 is borderline.

21 rules across four categories. Content: spam-trigger phrases, urgency language, generic greetings, forward-to-friends, 100% guarantees, money-back claims. Format: excessive caps, multi-exclamation, money-symbol runs, heavy emoji. Structure: image-only emails, white-on-white hidden text, tiny/zero-size fonts, missing unsubscribe link, missing physical address, oversized HTML. Links: URL shorteners, IP-based URLs, link/text mismatch (phishing pattern), excessive total links, heavy use of tracking redirectors.

True - transactional emails (receipts, password resets, account notifications) are exempt under CAN-SPAM. The rule fires anyway because the tool can't tell from the body alone whether the email is transactional or marketing. If your test input is a receipt and the no-unsubscribe rule fires, ignore it. The rule is meaningful for marketing/promotional sends, where missing it is a real deliverability hit.

It's POSTed to a Cloudflare Worker, scored, and the result returned. The Worker doesn't log, store, or correlate the input. The rules engine runs entirely in the Worker (no third-party calls), so the only entity that sees your HTML is Cloudflare's edge runtime. We chose this over a fully client-side implementation because the rules engine is ~30 KB - acceptable as a one-time download but better as a Worker call to keep the page fast.

No tool can. Real spam filters use machine-learning models trained on millions of emails, plus IP reputation, sender history, recipient engagement, and dozens of other signals you can't audit at design time. This tool catches the well-understood content-side issues. Treat it as a hygiene check, not a delivery prediction.

Common false positives: (1) marketing copy that legitimately uses urgency ("Last chance to register"), (2) a sale email with prices and money symbols, (3) a transactional email that doesn't need an unsubscribe link, (4) a newsletter pasted without the layout's unsubscribe footer. Look at WHICH rules fired - if they're all rules you've consciously decided are appropriate for your send, the low score is a false alarm.

The rule definitions are in the open-source min8t-tools repo at workers/api/spam-rules.js (MIT licence). Each rule has a name, category, weight, and test function. Read them, propose changes, or fork. The list will grow over time.