Words that trigger spam filters, and safer alternatives

Word choice is a minor signal, not the main one

For an established sender with a healthy reputation, no single word in an email body will flip a message into spam on its own - modern filtering weighs sender reputation, engagement history, and authentication far more heavily than keyword matching. For a new domain, a cold-outreach mailbox, or a sender already on the edge of a reputation problem, though, content patterns can tip a borderline decision, particularly in the subject line, which filters and humans both weigh disproportionately.

The categories that still matter

Urgency and pressure language - "act now," "expires today," "last chance" - correlates strongly with both spam and genuinely low-quality marketing, because legitimate time-limited offers are a small fraction of the mail that uses this language.

Unearned exclusivity and too-good-to-be-true framing - "you've been selected," "guaranteed," "no risk" - trips both automated filters and a reader's own spam instinct, often before the filter even gets involved.

Financial pressure and reward framing - "free money," "cash bonus," "double your," combined with dollar signs and all-caps - remains one of the more consistently flagged patterns because it's common in genuinely fraudulent mail.

Personal and privacy-probing language in cold outreach - referencing a recipient's personal life, health, or financial situation in a first-touch message from an unfamiliar sender reads as invasive, which drives complaints even when it technically isn't spam.

A quick reference

Subject lines matter more than body copy

A subject line is evaluated before the recipient (or, in some cases, the filter) has any other context, so it carries more weight per word than the same language buried in a paragraph of body copy. A body paragraph that mentions a deadline in passing is lower risk than the same word appearing in the subject line.

Test the pattern, not just the word

A single word rarely causes a problem on its own - it's usually several risk patterns compounding (urgency language, in a subject line, from a new domain, with a tracked link and an attachment) that pushes a borderline send into spam. If you're troubleshooting a filtered campaign, look at the combination of signals rather than scanning for one "bad word" to remove.

Why putting risky text into an image doesn't help

A tempting workaround for urgency or financial-pressure language is putting it into an image instead of plain text, on the theory that a filter scanning text content won't see it. This doesn't reliably work and often makes things worse: image-heavy, text-light emails are themselves a separate spam-correlated pattern, and many modern filters run optical text recognition on embedded images specifically because this workaround became common enough to need a countermeasure. The net effect is you've added the image-ratio risk on top of the original wording risk, rather than avoiding either one.

The more durable fix is rewriting the message so the underlying claim doesn't need risky framing in the first place - stating a real, specific deadline plainly instead of manufacturing urgency, or describing an actual offer without resorting to "guaranteed" or "no risk" language that reads as evasive to both filters and recipients. Keep a running internal list of phrases that have actually correlated with placement problems in your own sending history, rather than relying solely on generic published lists - the categories above are a reliable starting point, but the specific phrases that matter most are often ones tied to your own audience and content style. Revisit that internal list every time you run a seed-list placement test on new content, adding any phrase that correlated with a worse result even if it doesn't match a commonly published trigger-word category.

What not to do

Do not run every email through a rigid banned-word list and assume that alone fixes deliverability - it addresses a minor signal while ignoring reputation, authentication, and list quality, which matter far more. Do not strip all urgency language from genuinely time-sensitive, legitimate communications - a real deadline stated plainly isn't the problem; manufactured urgency is. Do not assume a filtered send was caused by wording before checking bounce rate, complaint rate, and authentication status first.

MailPilot's seed-inbox placement checks show you where a specific send actually landed across providers, so you can isolate whether a filtering problem is content-related or a reputation and authentication issue before changing your copy.