Spam Trigger Words Checklist: Words and Patterns That Trigger Spam Filters

Why Trigger Words Alone Rarely Cause Filtering

Spam filters in 2026 are reputation-first, content-second. Gmail, Outlook, and Yahoo all run content scoring as a final tiebreaker — the layer that decides between inbox and spam for a message from a sender with a borderline reputation. Authentication, list quality, and engagement history do most of the work; content only matters when other signals are ambiguous or negative.

The implication for senders: chasing a clean trigger-words score while ignoring authentication is wasted effort. The inverse is also true — fixing authentication while ignoring the worst content patterns caps your deliverability at the inbox-vs-spam boundary instead of solidly inside the inbox. For the broader deliverability playbook, see Cold Email Deliverability → and SPF, DKIM, and DMARC Explained →.

Subject-Line Red Flags

The subject line is the first content the filter scores. It carries roughly twice the weight of the body on a per-character basis because recipients see it first and report on it first. The patterns that consistently raise the score:

• All-caps subject lines. Entirely capitalized words score higher than mixed-case. One capital word in a normal subject is fine; a subject written entirely in caps consistently scores as promotional.
• Multiple exclamation marks or question marks. "Buy now!!!" or "Sale???" scores measurably worse than "Buy now" or "Sale" — the filter treats repetition of punctuation as a strong promotional signal.
• Money symbols. Multiple dollar signs ($$$$, $$$), euro symbols, or other currency-symbol repetition score as high-pressure promotional. One symbol in a normal sentence is fine; stacking them is not.
• Classic promotional phrases. The words "FREE", "ACT NOW", "LIMITED TIME", "100% GUARANTEED", "WINNER", and "CLICK HERE" all score as promotional when used as standalone subject lines. They are not banned — used in normal context within a longer subject they are fine; used as the entire subject they consistently filter.
• Deceptive subjects vs From: mismatch. When the subject promises something the From: header doesn't deliver ("Your account is locked" from a no-reply@marketing domain), the filter treats the message as phishing-adjacent. This is the single most common trigger for Microsoft moves to the Junk folder.

Body Red Flags

Body scoring is less aggressive than subject scoring — recipients see the subject first and report on it first — but body patterns still raise the score enough to matter when other signals are borderline.

• URL shorteners. bit.ly, tinyurl, t.co, ow.ly, and similar services obscure the destination domain. Filters treat shortened URLs as a strong promotional-and-phishing signal — both because they hide the destination and because spammers disproportionately use them. Replace every shortened link in your body with the full destination URL on your own domain.
• IP-address links. Links that point to a raw IP (e.g. http://192.0.2.1/...) score as high-risk. Every legitimate link should resolve to a domain with a valid HTTPS certificate.
• Image-heavy HTML. A message that is a single large image with no text alternative scores as promotional. Filters can't read image content, so they treat the message as opaque — and the absence of readable text raises the score. Always include a meaningful text/plain alternative alongside any HTML.
• Broken HTML. Malformed tags, unbalanced tables, and invalid CSS in the HTML body raise the score. Some receiving servers reject at SMTP time with a 5xx; others Junk the message silently.
• "Click here" with mismatched anchor. Anchor text that says one thing but links to another domain scores as deceptive. The visible link text and the href must match the same root domain.
• Risk-claim language. "No risk", "100% safe", "guaranteed returns", "zero effort", "winner", "congratulations you won" — all score as promotional when used as standalone claims. They are fine in context (a testimonial that quotes a customer saying "no risk" works) but stack as risk signals when used in a list of selling points.

Common-Content Red Flags

A third category of red flags sits between the body and the headers — patterns in how the message is composed, not what it says.

• Deceptive subjects. Subjects that preview as one thing in the inbox but the body delivers something different. Gmail and Outlook both downweight senders who do this consistently.
• Sender / From: mismatch. The friendly From: name shows a brand but the underlying address is a generic no-reply@. Filters treat this as a phishing signal; recipients do too.
• Missing unsubscribe for bulk mailers. Since February 2024, Gmail and Yahoo require a working List-Unsubscribe header on bulk senders. Missing it produces hard bounces against Gmail and Yahoo recipients and counts as a content-policy violation. The full checklist is in Gmail Bulk Sender Requirements →.
• Reply-to pointing to a different domain. Reply-to set to a domain unrelated to the From: domain scores as phishing. The reply-to must be on the same root domain as the From:, or on a domain the sender controls.

All-Caps and Excessive-Punctuation Risk

All-caps and excessive punctuation are amplifiers that turn a normal phrase into a high-risk one. The math is intuitive once you see it scored:

• "Free" — neutral. The word alone scores as descriptive, not promotional.
• "FREE" — slightly elevated. Capital letters raise the score modestly. Used once in a normal subject, fine; used throughout, promotional.
• "FREE!!!" — high risk. Caps plus multiple exclamation marks compound; the score climbs sharply. Filters treat this as a classic promotional pattern.
• "FREE!!! CLICK HERE NOW!!!" — max risk. Caps, multiple exclamation marks, and a CTA phrase stack. This pattern alone is enough to push the message into Junk on a borderline sender.

The safer rewrite is almost always shorter and reads better to a human. "Free" with no exclamation beats "FREE!!!" on every axis — filter score, recipient perception, and reply rate.

Glossary: High-Risk Phrase → Safer Rewrite

The mapping below covers the phrases that most consistently raise the filter score when used as standalone claims. The safer rewrites preserve the meaning and almost always read better to a human recipient.

Most of the safer rewrites are not just less risky — they are more credible. "Available through Friday" reads as a real offer; "LIMITED TIME OFFER" reads as marketing. Filter score and recipient trust move in the same direction here.

The Pre-Send Checklist

Run this checklist before every campaign send. It takes under 10 minutes and is the difference between inbox and spam on borderline senders.

1. Subject line is mixed-case. No all-caps. No multiple exclamation marks or question marks. No stacked currency symbols.
2. Subject and From: align. The brand the subject promises is the brand the From: address shows.
3. Every link resolves to your own domain. No URL shorteners. No raw IP addresses. Anchor text matches the destination page.
4. Image-to-text ratio is reasonable. At least 60% text. Every image has alt text. The HTML is well-formed.
5. A text/plain alternative is present. Not HTML-only. The plain-text version is meaningful, not a stub.
6. List-Unsubscribe header is present and one-click. Required for bulk senders since February 2024; tested and working. The full checklist is in Gmail Bulk Sender Requirements →.
7. No deceptive claims. No "no risk", "100% guaranteed", "winner", or "congratulations" used as standalone selling points. Specific, verifiable claims only.
8. Reply-to is on the same domain. Not a free-webmail address. Not an unrelated third-party domain.
9. Authentication passes. Run the free SPF/DKIM/DMARC check → and confirm all three resolve and align.
10. List is engaged. Recipients have opened or clicked within the last 90 days. Anything colder is on a re-engagement track, not the main campaign.
11. Sending IP has clean reputation. No active blocklist hits. SNDS filter result green. SenderScore above 80.
12. Test send to a Gmail and Outlook.com. Both should land in the Inbox within five minutes. Anything in Junk on a clean sender is a content problem; fix and re-test before the main send.