Search "spam trigger words" and you will find the same recycled list on every result page: FREE!, ACT NOW, LIMITED TIME, MAKE MONEY FAST, CLICK HERE, BUY NOW, and a few hundred more. The list has barely changed since 2008.
The problem is that modern spam filters do not work the way those lists assume. The "100 spam words to avoid" template is one of the most outdated artifacts in email marketing advice. This guide explains what 2026 filters actually evaluate, which specific words still matter (a short list), and why sender reputation overrides almost everything in your subject line and body copy.
- Modern spam filters (Gmail, Outlook, Apple Mail) use machine learning models that evaluate hundreds of signals together. No single word triggers a filter on its own.
- Sender reputation, authentication, engagement history, and recipient relationship account for the vast majority of filtering decisions. Word choice rarely outweighs these.
- The remaining word-level signals that matter are concentrated in patterns (excessive capitalization, exclamation density, all caps subject lines, urgency stacking) rather than individual vocabulary.
- Some words still carry meaningful filter weight, but only in specific contexts (financial offers, prescription drugs, sexual content, cryptocurrency).
- The most reliable way to avoid spam folders is to build a strong sender reputation, not to obsess over individual word choices.
The Myth of the Spam Word List
The "list of spam trigger words" concept comes from an earlier era of email filtering, when rule-based filters like SpamAssassin assigned point values to individual keywords. A message scoring above a threshold was classified as spam. In that world, "FREE" was worth 0.3 points, "ACT NOW" was worth 0.5, and stacking enough of them tipped a message into the spam folder.
That world ended around 2015 for major mailbox providers. Gmail, Outlook, and Apple Mail now run machine learning classifiers that evaluate hundreds of features simultaneously, including:
- Sender domain and IP reputation across millions of historical messages
- Authentication results (SPF, DKIM, DMARC alignment)
- Recipient engagement history with the sender
- Aggregate complaint rates from similar senders
- HTML structure, link patterns, image ratios
- Sending pattern anomalies and volume changes
- Content features including word choice, but as a small fraction of the overall score
Word choice still matters, but its weight has collapsed relative to sender behavior. A reputable sender can use the word "free" in a subject line every week without consequence. A new sender with no reputation can avoid every word on every spam list and still land in spam because they have no engagement history.
How Modern Spam Filters Actually Evaluate Words
To the extent words matter, they matter in combination and context. A modern filter does not say "the word FREE means +0.3 spam score." It evaluates the entire message holistically:
- Phrase patterns: "Click here to claim your free prize" hits multiple spam-correlated phrases in sequence. The model has learned that this exact phrase pattern correlates with bad mail in training data.
- Tone consistency: A legitimate newsletter has consistent voice and vocabulary. Spam often jumbles aggressive urgency, hype, and disclaimers in incoherent ways.
- Context misalignment: A B2B SaaS company suddenly using phrases like "your last chance" or "act before midnight" looks out of character. Filters notice when sender history does not match current content.
- Punctuation density: Three exclamation points in a six-word subject line is a strong signal. The exact words matter less than the punctuation pattern.
- Capitalization: ALL CAPS WORDS in subject lines or body remain a reliable spam correlate. A single capitalized word is fine; sentence-level capitalization is not.
The Few Words That Still Carry Real Weight
If you eliminate the noise from outdated spam word lists, the words that still carry measurable filter weight in 2026 cluster around regulated and frequently abused categories:
Financial offers
"Pre-approved," "no credit check," "cash bonus," "lowest rates," "consolidate debt," "easy income," "make money from home." Filters have learned these phrases correlate with predatory lending, MLM schemes, and financial fraud.
Prescription drugs and health claims
"Viagra," "Cialis," "weight loss miracle," "diet pill," "lose weight fast," and most pharmacy brand names trigger pharmaceutical spam classifiers. Even legitimate health newsletters need to be careful with these terms.
Sexual content
Explicit terms, dating service vocabulary, and adult content keywords trigger filter categories that are heavily trained on adult spam. Even with consent and engagement, these words carry meaningful weight.
Cryptocurrency and investment hype
"Crypto millionaire," "guaranteed returns," "ICO," "trading signals," "double your investment." Filters have been heavily trained on crypto scams since 2017 and continue to expand pattern recognition.
Urgency stacking
Words like "urgent," "expires," "act now," "last chance," "limited time," "deadline" carry minor weight individually but compound when stacked. One urgency word is fine. Three or four in a subject line plus body is a spam pattern.
Watch out: Words that are perfectly fine in isolation can become problematic in combination. "Free shipping on your order" is benign. "FREE!!! Limited time only - Act now or lose this URGENT offer FOREVER" stacks every problematic pattern simultaneously.
Patterns That Matter More Than Individual Words
The signals that actually move filter decisions are mostly structural rather than vocabulary-based.
Subject line capitalization
Subject lines in all caps or with multiple all-caps words signal spam patterns. Title case and sentence case are normal; SCREAMING IS NOT.
Exclamation point density
Multiple exclamation points in subject lines or body copy correlate strongly with spam patterns. One exclamation is fine. Three or more is a signal.
Subject line and body mismatch
Bait subject lines that promise something the body does not deliver are flagged through engagement signals. Users delete or report these messages, training the filter against the sender.
Excessive HTML obfuscation
Hiding text inside images, using zero-pixel fonts, white-on-white text, or unusual character substitution (like 0 instead of o) are classic obfuscation techniques. Filters have been trained on them for two decades and still catch them.
Heavy image-to-text ratio
Mostly-image emails with minimal text have historically been used to evade text-based filters. Modern filters treat low text-to-image ratios as a moderate spam signal.
Spammy link patterns
URL shorteners, link redirects through unknown domains, mismatched display text and target URLs, and links to known-bad domains all carry weight. The links matter more than the words around them.
Sender Reputation Overrides Almost Everything
The single most important variable in whether a message lands in the inbox is who is sending it, not what the message says. A sender with a strong domain reputation, valid SPF, valid DKIM, aligned DMARC, and consistent engagement history can use language that would land a new sender in spam.
This is why obsessing over individual word choices is usually misallocated effort. Higher leverage activities include:
- Build and maintain authentication. Verify with a DMARC checker and SPF checker.
- Maintain clean lists. Use real-time email verification on signup and prune unengaged subscribers regularly.
- Monitor sender reputation using Google Postmaster Tools and Microsoft SNDS.
- Keep complaint rates below 0.1 percent and bounce rates below 2 percent.
- Segment for engagement and send relevant content to recipients who actually want it.
If your emails are going to spam and you have a strong sender reputation, the cause is almost never a single word in your subject line. Audit your authentication, list hygiene, and engagement segmentation before rewriting your copy. The words are usually the last 5 percent of the problem.
What Testing Tells Us
Independent placement testing across thousands of campaigns has converged on a consistent finding: word-level changes rarely move placement when sender reputation is constant. Swapping "free" for "complimentary" or "act now" for "available today" produces no measurable inbox placement difference for established senders.
Where testing does show movement:
- Removing all-caps subject lines: meaningful improvement
- Reducing exclamation points to one or zero: small improvement
- Adding plain text alongside HTML: small improvement
- Reducing image-to-text ratio: small improvement
- Removing URL shorteners: meaningful improvement
- Adding valid DKIM: large improvement
- Building domain reputation over 90 days: largest improvement
The pattern is clear: structural choices matter, vocabulary mostly does not, and infrastructure matters more than either.
Most "spam word checker" tools score messages based on rule sets that were last meaningfully updated in 2014. They are essentially measuring how a 2014 SpamAssassin instance would have scored your mail, which has limited correlation with how Gmail or Outlook will treat it in 2026.
A Modern Content Checklist
Rather than chase a list of forbidden words, follow these structural guidelines:
- Use sentence case in subject lines. Avoid all caps and multi-word capitalization.
- Limit exclamation points to one maximum, ideally zero.
- Match subject line tone to body content. No bait and switch.
- Maintain a text-to-image ratio of at least 60 percent text by content area.
- Use plain English links with target URLs that match the display text.
- Avoid URL shorteners in marketing email. Use your own tracking domain if needed.
- Include a working unsubscribe link and List-Unsubscribe header.
- Use the words your audience actually uses, not vocabulary engineered to dodge filters.
Frequently Asked Questions
The words that still carry measurable filter weight cluster around regulated categories: prescription drug names, financial fraud vocabulary (pre-approved, no credit check, easy income), cryptocurrency hype (guaranteed returns, trading signals), and explicit sexual content. Most legacy "spam word lists" are outdated and the specific words on them rarely move placement for established senders.
No. The word "free" by itself does not trigger modern spam filters. Established senders use it routinely without consequence. What triggers filters is the combination of "free" with all caps, multiple exclamation points, urgency stacking, and other classic spam patterns. The word is not the problem; the structural pattern is.
Modern filters from Gmail, Outlook, and Apple Mail use machine learning models that evaluate hundreds of features at once. The largest weight goes to sender reputation, authentication results, and recipient engagement history. Content features including word choice account for roughly 15 percent of the decision. Word choice rarely moves placement when sender reputation is constant.
Excessive exclamation points correlate with spam patterns. One exclamation point is fine; three or more in a subject line is a measurable spam signal. The density and combination matter more than the presence of any single punctuation mark. Sentence case and minimal punctuation perform best across all major mailbox providers.
Most spam word checkers use outdated rule sets that approximate 2014-era SpamAssassin scoring, which has limited correlation with how modern filters at Gmail, Outlook, and Apple Mail actually work. They are not harmful but the time is better spent on authentication, list hygiene, and engagement monitoring, all of which carry far more weight in real placement decisions.