Spam Trigger Words Are Misunderstood

The Word List Is a Trap

Every spam trigger word article on the internet tells you the same thing. Avoid "FREE." Don't write "ACT NOW." Stay away from "GUARANTEED."

That advice is incomplete to the point of being useless for cold email senders in the current environment.

The cold email senders who are losing the inbox battle right now are not losing because they wrote "limited time offer." They're losing because modern spam filters have moved far beyond keyword matching. Filters now score behavioral patterns, copy uniformity, sender history, domain naming conventions, engagement velocity, and AI-generated prose structure - all in real time, all before your email hits one single inbox.

Knowing which words to avoid is table stakes. What you need to understand is the full picture of what gets you flagged today - including the triggers that appear on no word list anywhere.

This article covers both. Every major word category you need to know. Plus the non-obvious triggers that are killing campaigns right now.

How Modern Spam Filters Work

Old spam filters worked like a checklist. Email contains "FREE MONEY" - flag it. Email contains "GUARANTEED" - flag it. Simple keyword matching against a static list.

That model is mostly dead.

Today's filters at Gmail and Outlook run machine learning models trained on billions of emails and the actions recipients take after seeing them. Google's Resilient and Efficient Text Vectorizer (RETVec) can detect deliberately misspelled and manipulated text - things like "fr€€ w1nn3r" or "V!agra" - that traditional filters completely miss. RETVec helped Gmail detect 38% more spam while reducing false positives by 19.4%. Gmail also uses TensorFlow models that block an additional 100 million spam messages per day.

Microsoft's Exchange Online Protection runs its own AI layer that performs linguistic analysis, structural analysis, and behavioral analysis simultaneously. It is checking your language and tone, your message headers and URL structures, and whether your sending patterns deviate from what a normal sender looks like.

The practical implication for cold email senders is this: spam filters now evaluate system-level patterns across your domain reputation, send behavior, and engagement signals. A single spam word won't kill you. A spam word in combination with a new domain, a mass-send pattern, identical copy across 500 sends, and low engagement will absolutely kill you.

Think of it like a credit score. One late payment doesn't disqualify you from borrowing. But when late payments stack alongside high utilization, short credit history, and new accounts - the system moves against you fast. Language works the same way. The same phrase that lands fine from a trusted domain with strong engagement history can send you straight to spam when it comes from a new domain at volume.

One critical reality: cold email faces the highest scrutiny because it starts without any trust. No opt-in. No engagement history. Small language choices carry disproportionate weight, especially when combined with new domains or rising send volume.

The Spam Complaint Threshold You Need to Know

Before we get into specific words, understand what's at stake with your numbers.

Gmail starts filtering your emails when your spam complaint rate hits 0.3%. To stay completely safe, keep your rate under 0.1%. That's one to two spam complaints per 1,000 sends. Just one or two people hitting "report spam" out of a thousand recipients can begin degrading your domain.

41% of new domains get flagged within the first 30 days of cold outreach. That is not a small number. The combination of a fresh domain, no sending history, and aggressive volume creates a risk profile that filters treat as suspicious regardless of your copy.

Once your sender reputation drops, rebuilding it can take months. Prevention is dramatically easier than recovery. A damaged domain can take three to six months to rehabilitate. Some never fully recover.

Keep bounce rates under 1%. Maintain open rates above 65% per week on your sending domains. These are the operational numbers that determine how much scrutiny your copy will face. A high-reputation domain gets the benefit of the doubt on borderline language. A low-reputation domain does not.

The Classic Spam Trigger Word Categories (With Safe Replacements)

These are the categories that have always mattered and still do - but they matter more or less depending on everything else about your send. A clean domain with strong engagement can tolerate occasional borderline language. A new domain with weak warmup cannot afford any of it.

Financial Promises and Get-Rich Language

Filters have been trained on financial scam emails for decades. Any phrase that implies easy money, passive income, or financial shortcuts triggers immediate scrutiny.

Avoid: "earn extra income," "make money fast," "financial freedom," "instant cash," "get paid," "passive income," "earn from home," "extra income," "unlimited income," "work from home," "cash bonus," "big bucks," "money back," "make $," "profit," "get rich."

Safe replacements: Talk about outcomes specific to the prospect's business. "Reduce churn" instead of "make more money." "Shorten your sales cycle" instead of "increase revenue fast." "We helped a company in your space go from 120 to 340 qualified calls per quarter" hits differently than "boost your income."

Urgency and Pressure Tactics

Urgency words are high-risk in cold email specifically. The reason is context. In a newsletter from a brand you've subscribed to, "limited time offer" is a normal commercial phrase. In a cold email from a domain you've never seen before, the same phrase reads as manipulation - filters have learned to treat it that way.

Avoid: "act now," "urgent," "limited time," "offer expires," "last chance," "now only," "deadline," "don't delay," "respond immediately," "time sensitive," "hurry," "this won't last," "while supplies last," "final notice," "one time only."

Safe replacements: If you have a genuine time constraint, state it factually. "We have three onboarding slots open in Q3" is specific and credible. "Act now before it's too late" is a red flag - to both filters and humans.

Exaggerated Claims and Guarantees

Filters flag guaranteed outcomes and miracle claims as scams and deceptive marketing. This applies even if your claim is technically true. The language pattern carries the risk regardless of reality.

Avoid: "guaranteed," "risk-free," "100% free," "no cost," "promise you," "satisfaction guaranteed," "zero risk," "100% satisfied," "certified," "as seen on," "best price," "amazing," "incredible," "unbeatable."

Safe replacements: Replace guarantees with specifics. "Our median customer sees X result in Y timeframe based on Z clients" is more credible and far less risky than "guaranteed results."

Aggressive Sales and Promotional Language

This is the category most cold email senders accidentally fall into. These are normal marketing phrases that feel natural to write - and that carry real risk in subject lines.

Avoid in subject lines and opening sentences: "exclusive offer," "special promotion," "buy now," "order now," "click here," "discount," "sale," "free trial" (in certain contexts), "deal," "opportunity," "save big," "lowest price," "best deal," "check it out," "don't miss out."

The placement matters. Words appearing in subject lines and opening sentences are scrutinized more than the same words buried in the body. A spam word in your subject line is a much bigger signal than the same word in your third paragraph.

Subject Line-Specific Risks

Subject lines get more scrutiny than body copy.

A few specific patterns are particularly dangerous:

Using "Re:" or "Fwd:" in a subject line when there was no prior conversation. Filters know this is a manipulation tactic and flag it. So do humans - and human spam reports feed back into filter training.

ALL CAPS in subject lines. This pattern is so strongly associated with spam that it carries risk even in otherwise clean copy.

Multiple exclamation marks. "Big news!!!" reads as spam formatting to both humans and filters.

Misleading or deceptive subject lines. If your subject line implies something your email doesn't deliver, you're creating both a filter risk and a complaint risk. Recipients who feel tricked click "report spam."

What works instead: lowercase, conversational subject lines. One operator tracked across thousands of sends and found that subject lines including the recipient's company name drove 35% more opens than generic lines. Specificity signals human effort. Specificity is the opposite of what spam looks like.

The Non-Obvious Triggers Nobody Puts on Their List

Deliverability articles give you the standard word categories and stop. Triggers exist that have nothing to do with classic spam vocabulary - and these are the ones quietly destroying campaigns right now.

Large Numbers in Email Body Copy

Writing a number like "100,000" in your email body is a documented spam trigger. This applies to big round numbers in financial contexts - "reach 100,000 prospects," "we've helped 50,000 companies," "generate $500,000 in pipeline." Spam filters associate large numeric claims with the financial promise language that populates scam emails. The number itself becomes a risk signal even when the surrounding copy is clean.

The fix is simple: round down to a smaller illustrative number, use a word instead of a numeral for large figures, or reframe the claim. "Thousands of B2B teams" instead of "10,000 customers." "Generated over half a million in new pipeline" instead of "$500,000."

Domain Names That Scream "Outreach"

Your domain name itself is a spam signal - not just your sending history.

One operator running 350 active domains across 40+ clients ran a clean A/B test on domain naming conventions with the same DNS setup, same warmup protocol, same copy. Naturally named domains - the kind that looked like real company websites (tryacme.com, acmehq.com) - hit 86% inbox placement. Systematically patterned domains - the kind that look like sending infrastructure (revio-mail1.com, outbound-revio.com) - hit 71% inbox placement. Same everything except the name. Inbox placement dropped 15 points based on domain naming alone.

Google specifically penalizes domains with words like "mail," "send," "outreach," or "reach" in the domain name itself. These words are heavily associated with bulk sending infrastructure. When your sending domain looks like a tool rather than a company, filters treat it accordingly.

Use domains that look like real business websites. A domain like getharveyco.com lands better than harvey-mail1.com. The domain is part of your identity as a sender. Make it look human.

UTM-Tracked Links in Cold Emails

This one is counterintuitive because UTM tags are standard practice in marketing. But in cold email, they are a deliverability liability.

UTM-tagged URLs look like this: yoursite.com/page?utm_source=cold&utm_medium=email&utm_campaign=q3. To a spam filter, that appended parameter string is a strong signal that the email is part of a mass campaign. Bulk marketing behavior is exactly what it is. Using UTM tags in cold email essentially self-labels your message as a broadcast, not a conversation.

The fix: use clean landing page URLs in cold email. Track campaign performance through reply rates and call bookings rather than click tracking. If you must track, use a redirect through a clean branded domain - not a raw UTM string.

Identical Copy Patterns Across Mass Sends

Modern filters detect when the same body copy, the same subject line, the same CTA phrasing, and the same signature format appear across hundreds of sends - even if each individual word is perfectly clean.

The trigger is the pattern. Think of it this way: if you see the same sentence structure across 500 emails, you don't need to know that sentence is "spammy" to know those emails came from an automated system. Filters think the same way.

This is why spintax is one of the highest-ROI deliverability tactics in cold email right now. Spintax means creating variation within your template so each send is slightly different - different subject line phrasing, different opening sentence, different CTA. Without spintax running at scale, up to 75% of outreach can land in spam from pattern detection alone, even with clean vocabulary and solid domain reputation.

Build variation into every element: subject lines, opening sentences, the body copy itself, the CTA, and the sign-off. Even small differences break the pattern signal that flags your send as automation.

Tracking Pixels in Cold Emails

Open tracking via a tracking pixel is another marketing-standard practice that creates deliverability risk in cold email specifically. A tracking pixel is a tiny invisible image loaded when the email is opened. Spam filters know what they are. Their presence signals: this is a monitored mass send, not a real conversation.

For cold email, send plain text without tracking pixels. You lose the ability to track opens individually - but you gain inbox placement. The tradeoff is clear when you consider that an email in the spam folder has a 0% open rate regardless.

Very Short Emails and Outlook

Here is a counterintuitive finding specific to Outlook. Shorter emails inbox better with Microsoft's filters, which weight content length as a signal of human vs. automated sending. The logic makes sense - real individual emails tend to be short. Bulk marketing emails tend to be long and HTML-heavy.

One practitioner documented a specific insight about a widely repeated claim in the cold email community: that emails under 30 words will "definitely" land in Outlook's inbox. The nuance is that this is directionally true - shorter equals more human-seeming to Outlook's filters - but a 30-word hard cutoff is an oversimplification. The actual dynamic is that Outlook penalizes heavy, long, HTML-formatted emails more aggressively than Gmail does. Keeping cold emails under 100 words outperforms longer emails by roughly 3x in reply rates in most tested scenarios. Brevity helps deliverability with Outlook.

Plain text. Short sentences. No HTML formatting tricks. No bold text or colored fonts. These are not just stylistic recommendations - they are filter signals. HTML-heavy formatting is a broadcast pattern. Plain text is a conversation pattern.

The "Unsubscribe" Paradox

CAN-SPAM law requires commercial email to include an unsubscribe option. But for cold email specifically, including a standard unsubscribe link creates a problem: spam filters see a mass-formatted unsubscribe footer and classify the email as a marketing broadcast, not a personal outreach.

The workaround used by experienced cold email operators: replace the word "unsubscribe" with a conversational opt-out line written in plain text. Something like "Reply with 'not interested' if you'd like me to stop reaching out" achieves legal compliance without triggering the broadcast signal that the word "unsubscribe" carries. Some go further and use spintax on the opt-out phrasing - rotating between different plain-text variations - so even the opt-out line doesn't look identical across thousands of sends.

AI-Generated Copy Is Now a Spam Signal

This is the deliverability story dominating the current moment. AI writing tools have made average outreach copy more uniform, more predictable, and more pattern-heavy than ever before. And email providers have trained their filters on billions of these AI-generated emails.

Gmail and Outlook are now actively detecting AI-written cold emails. They recognize the structure. Filters have trained on the patterns. If your email has high "AI perplexity" - meaning it reads too much like generated text - it routes straight to Promotions or Spam.

The tells are well-documented at this point. "I hope this email finds you well." "I was impressed by [Company Name]'s recent growth." "I'd love to explore synergies." "I wanted to reach out because..." These patterns appear across millions of AI-generated cold emails. Filters recognize them the same way a human would recognize a form letter.

Large language models have made outreach copy more uniform across the board. Filters increasingly recognize AI-shaped phrasing - not as an explicit penalty in isolation, but as a signal that folds into low-trust sender behavior when it's overused or poorly adapted to the specific recipient.

The practical implication: generic AI output is a liability. Human-sounding, specific, slightly imperfect copy outperforms polished AI prose in deliverability. Write like you text. Lowercase subject lines. Short sentences. Specific references to the recipient's actual business. A genuine postscript that proves you read something about this person.

One documented signal: AI rarely writes a natural P.S. effectively. A human postscript that references something real about the recipient - "PS - saw you just closed a Series A, congratulations" - signals to both the filter and the human that this email came from a real person who did actual research.

The reply rate data shows the stakes. Cold emails needed to generate one positive reply went from roughly 120 sends in early multi-year tracking to 200, and then to 430. The volume required has more than tripled. Google and Microsoft are the gatekeepers of every B2B inbox, and AI detection is now a core part of their filtering logic.

The Domain Lifecycle Problem

I see this constantly - cold email senders treating domains as permanent infrastructure. They're not. Every sending domain has a lifecycle, and understanding that lifecycle changes how you think about spam trigger words entirely.

Here is the pattern documented by practitioners running large-scale cold outreach across hundreds of domains: micro-degradation of inbox placement begins around weeks 10 to 14 of active sending, then accelerates quickly. By the time you notice the deliverability drop, you're already behind.

Proactive domain rotation - retiring domains before degradation hits, not after - is one of the most effective deliverability tactics available. It outperforms any copy tweak or word substitution you can make.

The sending pattern within a domain's lifecycle also matters. "Bursty" sending - human-like clusters of emails rather than perfectly even intervals - produces 6% better inbox placement than randomized or evenly spaced sending. This was tested across 2,000 mailboxes over six weeks. At 50,000 daily send volume, 6% better inbox placement means roughly 3,000 more emails hitting the primary inbox every single day.

Reply velocity in the first two hours of a campaign also shapes how email service providers treat your entire send batch. Early positive signals - replies, forward actions - tell the algorithm this is content people want. If your first 200 sends go cold with no engagement, the algorithm uses that signal to treat the remaining sends more harshly.

A Practical Cold Email Deliverability Checklist

Here is what the best practitioners in cold email are doing right now, distilled into a checklist you can work through before any campaign goes out.

Domain setup and naming

• Domain name looks like a real company website, not sending infrastructure
• No words like "mail," "send," "outreach," or "reach" in the domain name
• SPF, DKIM, and DMARC properly configured (these are now mandatory - providers auto-filter senders without them)
• New domains fully warmed for at least four to six weeks before sending cold outreach
• Active monitoring for domain degradation - plan to rotate before week 10 to 14

List quality

• Email list verified before sending - bounce rate under 1%
• No purchased lists without verification
• No role-based addresses (info@, support@, contact@)
• No disposable or invalid emails

Copy

• No HTML formatting - plain text only
• No tracking pixels
• No UTM-tagged links
• No large round numbers in the body ("100,000", "$500,000")
• No ALL CAPS in subject lines or body
• No multiple exclamation marks
• No urgency phrases ("act now," "limited time," "respond immediately")
• No financial promise language
• No guaranteed outcome claims
• Spintax applied to subject line, opening sentence, body, CTA, and sign-off
• "Unsubscribe" replaced with conversational plain-text opt-out
• Subject line includes recipient's company name or specific reference

Sending behavior

• No more than 30 cold emails per inbox per day
• Bursty, human-like sending intervals (not evenly spaced automation)
• Spam complaint rate actively monitored - stay under 0.1%
• Open rate benchmark maintained above 65% per week
• Email length kept under 100 words where possible

Safe Replacements for the Highest-Risk Phrases

Here is a quick reference for substituting the phrases that consistently cause problems in cold email copy.

The Cold Email vs. Email Marketing Distinction

Almost every spam trigger word guide on the internet is written for email marketing - newsletters, broadcast campaigns, opted-in lists. Cold email operates under different rules and different filter treatment.

For email marketing, certain promotional language is expected and tolerated because the filter context includes the opt-in relationship. A subscriber who signed up for a deal newsletter expects promotional language. The filter knows this.

For cold email, there is no opt-in history, no engagement baseline, and no pre-existing relationship. Every element of your email is evaluated cold. A phrase that would pass cleanly in a newsletter can tip a cold email into spam because the same words carry a different signal weight in a different context.

This is why generic spam word lists fail cold email senders. The lists were built for email marketing. They don't account for the elevated scrutiny cold email faces or the behavioral signals that matter specifically for cold outreach - domain naming, send patterns, copy uniformity, engagement velocity.

Cold email runs on completely different rules. Treat it differently.

Testing Your Copy Before You Send

The best practitioners don't guess - they test. Before launching a campaign, run your email through a spam score checker that mirrors what Gmail and Outlook use. I've used nearly every cold email tool out there - almost all of them have this built in.

Send test emails to seed accounts across Gmail, Outlook, and Yahoo before full deployment. Inbox placement can vary significantly across providers. An email that lands in Gmail's primary inbox might land in Outlook's spam folder if the copy and sending infrastructure aren't dialed in. Sending 5,000 emails blind is a wasted week.

Test subject line variations against each other. The data is consistent: specific, lowercase, personalized subject lines outperform generic promotional ones by a wide margin. "Quick question about [Company]'s outbound" outperforms "Exclusive offer for [Company] today" every time - in deliverability and in human response.

If you want to build a list worth emailing in the first place, Try ScraperCity free - it lets you search millions of B2B contacts by title, industry, location, and company size, and includes a built-in email verifier to keep your bounce rate under that 1% threshold from the start.

What to Do If You're Already in Spam

If your inbox placement has already dropped, the recovery path is methodical, not fast.

Stop sending from the damaged domain immediately. Do not try to "push through" a deliverability problem with more volume. More volume with a damaged domain accelerates the problem.

Start warming a fresh domain in parallel. This takes four to six weeks minimum before you can resume cold outreach. Use that time to audit your copy against everything in this article, restructure your domain naming convention, and rebuild your list quality with verified contacts.

Some practitioners quarantine damaged domains for 30 to 60 days of warmup - no cold sends, only reply activity and engagement - before trying to restore them. Results vary. A domain with a deep spam complaint history may never fully recover. Prevention is always cheaper than rehabilitation.

Monitor your sending metrics obsessively going forward. Spam complaint rate under 0.1%. Bounce rate under 1%. Open rate above 65% per week. These numbers determine whether your next campaign will reach an inbox at all.

The Bigger Picture

The old model of cold email deliverability was: don't write "FREE" in your subject line and you'll be fine.

The current model is: every element of your send - your domain name, your sending pattern, your copy uniformity, your list quality, your engagement velocity, the structure of your prose, the presence or absence of tracking pixels - contributes to a continuous reputation score that determines where every email you send goes.

Spam trigger words still matter. Filters penalize them. But they are one layer of a much larger system. The senders who treat word avoidance as the whole solution are the same ones watching their reply rates decline year over year without understanding why.

The senders who understand the full picture - who treat deliverability as infrastructure, not copywriting - are the ones who still have functioning cold email programs today.

Build your system around the full picture. The word list is just the beginning.