Definition

Email spam filtering is the automated process by which mailbox providers analyse incoming messages and decide whether to deliver them to the inbox, route them to the spam folder, or reject them entirely. The filtering process evaluates hundreds of signals across four major categories: content analysis (the words, HTML structure, image-to-text ratio, and links in the message), authentication (SPF, DKIM, and DMARC pass/fail status), sender reputation (domain and IP address history, complaint rates, and bounce rates), and recipient engagement patterns (how the specific recipient and recipients like them have interacted with similar messages in the past). Gmail alone evaluates 100+ signals per message in under 100 milliseconds.

Spam score is a numerical representation of how likely a message is to be classified as spam, typically calculated by third-party testing tools rather than mailbox providers themselves. Tools like Mail-Tester, Litmus Spam Testing, and SpamAssassin assign a score based on the aggregate weight of triggered spam rules. A score under 5 is generally considered safe, 5-10 is risky and may trigger spam placement with certain providers, and scores above 10 indicate the message is likely to be blocked or routed to spam by most providers. Spam scores are diagnostic tools with imperfect correlation to actual inbox placement — a perfect spam score does not guarantee inbox delivery, and a moderate score does not guarantee spam placement.

Common spam triggers include excessive use of spammy trigger words ("free," "guaranteed," "act now," "limited time," "click here"), high image-to-text ratio (over 60% image content signals spamming behaviour), broken HTML or hidden text (text matching the background colour, CSS display:none on large blocks), suspicious link patterns (too many links, redirect-heavy URLs, or links to recently registered domains), and missing or improper authentication headers. Modern filtering systems are far more sophisticated than simple keyword matching — they use machine learning models trained on real user behaviour signals.

Best Practices

Get your authentication stack right before anything else. Publish SPF, DKIM, and DMARC records. Set DMARC to p=reject once your authentication pass rate is above 95%. Without proper authentication, your emails have no verified identity, making them indistinguishable from phishing messages in the eyes of mailbox providers. This is the single highest-impact spam reduction tactic available.

Monitor your spam complaint rate in Google Postmaster Tools and Yahoo Sender Hub. Keep the complaint rate under 0.1% of delivered emails (the widely accepted threshold used by Gmail). If your complaint rate exceeds 0.1%, Google may throttle or block your delivery. Check Postmaster Tools weekly for complaint rate, domain reputation, authentication pass rate, and feedback loop data.

Maintain an image-to-text ratio under 40%. Emails that are mostly images with minimal text are frequently flagged as spam because spam engines cannot analyse image content and legitimately informative emails include substantial text. If your design requires heavy imagery, include enough body text to bring the ratio under 40% and add meaningful alt text to images.

Avoid link patterns that trigger spam filters. Limit the number of unique domains in your email to 3-5. Use a single primary tracking domain. Avoid URL shorteners entirely. Ensure all links use HTTPS. Check that the link domains match the sending domain's ownership or email provider's tracking infrastructure. Suspicious link patterns are among the highest-weight spam signals.

Test every campaign through a spam testing tool before sending to your full list. Run the final HTML through Mail-Tester (free, sends to a test inbox and returns a score 0-10), Litmus Spam Testing (paid, tests against 16+ spam filters), or GlockApps (paid, tests major mailbox providers including Gmail, Outlook, Yahoo). Fix any issues scoring above 5 before proceeding to full send.

What is a good spam score?

Under 5 is safe for most mailbox providers. 5-10 is risky — messages may land in spam for some recipients depending on their past engagement with your domain. 10+ is blocked or consistently filtered to spam. Spam scores are indicative, not definitive — actual inbox placement depends on recipient-specific factors that spam tests cannot replicate.

Do spam trigger words still matter in modern filtering?

Yes, but less than they used to. Modern filters use context-aware machine learning rather than simple keyword blocking. Sending a "Free" offer in context with "shipping" may not trigger a filter if your reputation is good. However, combining multiple trigger words with high complaint rates is a strong negative signal. Write natural copy that avoids obvious spam tropes.

How does Gmail decide what goes to spam?

Gmail uses a machine learning model that evaluates 100+ signals including authentication, sender reputation, recipient engagement (opens, clicks, replies, spam reports), content analysis, and user-specific patterns. Gmail also uses "engagement filtering" — if most of a sender's recipients do not open or click their emails, Gmail assumes the content is not wanted and filters accordingly.

Can I be marked as spam even with perfect authentication?

Yes. Authentication proves you are who you say you are, but it does not guarantee inbox placement. If recipients mark your emails as spam, or if your engagement rates are very low, Gmail and other providers will still route your mail to spam. Authentication is a baseline requirement, not a deliverability guarantee.

What is the most common cause of spam placement?

Low recipient engagement is the most common cause for established senders. Spam complaints are the most common cause for new senders. For new senders, sending too much volume too fast without prior reputation (sending 100,000 emails on day one from a new IP) is the most frequent root cause of spam placement.