Spam trigger words: what word lists actually do to your email (and what filters weigh instead)

Somewhere in your company's shared drive there is probably a spreadsheet titled something like "words to never use in email": hundreds of entries, no source cited, copied from a blog post that copied it from an older blog post. This guide exists to retire that spreadsheet honestly. Word lists are not a modern description of how spam filtering works; they are a fossil of how it worked two decades ago, with a thin layer of remaining truth that is worth stating precisely.

Where the lists came from

Early-2000s spam filters really did score words. SpamAssassin, the open-source filter that shaped the era, shipped hundreds of content rules, each adding points toward a spam threshold, and commercial filters worked similarly. In that world, phrases of the classic type (historical examples, not a current banned list: "FREE!!!", "act now", "100% guaranteed", "no obligation", "earn money fast") genuinely carried scoring weight, and lists of them circulated as practical advice.

Filtering then moved on, for a simple reason: word rules are trivial to evade ("F R E E", images instead of text) and punish legitimate senders who happen to sell things. Google's own descriptions of Gmail filtering, and the sender guidelines it now enforces, are about authentication, sender reputation and user feedback. The word "free" appears in millions of wanted newsletters every day, and they arrive in the inbox because the sender's identity and history vouch for them; the mechanics are walked through in the Gmail spam filter guide.

Why the myth survives

If word-lists stopped describing reality years ago, why does every marketing team still have the spreadsheet? Partly because they are auditable: you can check a draft against a list in thirty seconds, while "improve your domain reputation" is a quarter-long project with no checkbox. Partly survivorship: any campaign that avoided the listed words and reached the inbox looks like confirmation, even though the authentication and reputation behind it did the work. And partly because the advice is cheap to republish, so each generation of blog posts copies the previous one's list without re-testing the premise. The result is effort spent on the one input filters weigh least, while the inputs they weigh most, verifiable identity and recipient behavior, go unattended. The sections below are those inputs, in the order the evidence supports.

What filters weigh instead

Modern placement decisions are dominated by evidence about the sender, not vocabulary:

• Authentication. Does the mail pass SPF, DKIM and DMARC for the visible From domain? Unauthenticated mail is filtered or rejected regardless of how carefully it is worded; verify yours with the email authentication checker.
• Reputation. The accumulated history of your domain and sending IP: past spam reports, bounce rates, blocklist appearances.
• Engagement and complaints. Whether recipients open, reply, delete unread, or hit "report spam." Google's published requirement holds bulk senders under a 0.3% spam complaint rate, a number about recipient behavior, not word choice.
• List quality. Mail to dead and unengaged addresses marks a list nobody consented to; list hygiene shows up in every reputation signal above.

Content still participates, but as one signal among many and mostly in combination: the same discount-heavy subject line that sails through for a domain with years of good history can tip the balance for a new, unauthenticated sender with no reputation to spend. That ranking of evidence, and why content sits last in it, is the spine of why emails go to spam.

The truth that remains in content

Retiring the spreadsheet does not mean content is irrelevant. A few content-level patterns still correlate with filtering, less as "trigger words" than as structural resemblance to actual spam:

• Deceptive mechanics. Misleading subject lines, link text that lies about its destination, sender names impersonating brands. These are also what the bulk-sender rules explicitly prohibit, and they generate the complaints that do the real damage.
• Spam-shaped structure. A single large image with almost no text, walls of exclamation marks and full-caps, dozens of links to low-reputation domains, URL shorteners hiding destinations.
• Manufactured urgency at scale. Not any one phrase, but the whole register of pressure ("final notice", "act immediately", again as examples of a style, not a list to regex against) reads as spam because actual spam leans on it so heavily.

Notice that every item is a pattern a human would also recognize as spammy. That is the honest modern rule: write like a business a recipient trusts, because the filter is, in effect, modeling that recipient's judgment.

What to do instead of word-policing

• Authenticate fully and verify it, per SPF, DKIM and DMARC explained, and run the email deliverability score for the DNS-side sweep.
• Mail people who asked. Double opt-in at the front, engagement-based sunsetting at the back.
• Make leaving easy: a working unsubscribe, one-click where the requirements demand it, keeps annoyance out of the complaint channel.
• Watch outcomes, not vocabulary: complaint rate in Postmaster Tools, bounce rate, and placement itself, seed-tested per how to check if emails go to spam.
• If campaigns are landing in spam right now, run the ordered diagnostic in marketing emails going to spam rather than rewriting copy first.

The full standing routine is in email deliverability best practices; none of its items is a word list.

Related guides

• Why emails go to spam — the evidence hierarchy filters actually apply
• Gmail spam filter — gate-by-gate mechanics at the largest mailbox provider
• Gmail sender requirements — the published rules that replaced word-lore
• Marketing emails going to spam — diagnosis when campaigns specifically are filtered
• Email authentication checker — verify the signals that outweigh vocabulary
• How to prevent emails from going to spam — the prevention routine, content included in its place

About this guide

Written by InboxRatio Editorial. Historical filtering behavior is described from SpamAssassin's public rule sets and contemporaneous documentation; current filtering weight is described from Google's and Yahoo's published sender requirements and filtering documentation. Example phrases are labeled as historical examples, not a compliance list. No vendor sponsorship influences it.

Methodology

InboxRatio's benchmark sends real campaigns through the platforms we review to a controlled seed list and records where each message lands. This guide draws on provider and project documentation rather than our measurements; the protocol is in how we test and our source rules in sources.

Last updated

20 September 2026. Provider filtering documentation reviewed quarterly.

Frequently asked questions

Do spam trigger words still exist? Not as a reliable mechanism. Word scoring was real in early-2000s filters, but modern mailbox providers weigh authentication, sender reputation and recipient engagement far more heavily. Content matters mostly in combination with weak sender signals, and as overall spam-shaped structure rather than individual words.

Will the word "free" send my email to spam? On its own, no. Millions of authenticated, well-regarded senders use it daily and reach the inbox. For a sender with no reputation, no authentication and a pushy all-caps layout, it is one more small weight on the wrong side of the scale, but the scale itself is built from sender evidence.

Why did my email land in spam if I avoided all the trigger words? Because vocabulary was never the deciding input. The likely causes, in order: authentication failures, weak or damaged sender reputation, poor list quality and engagement, then content structure. Work through the ordered diagnostic in why emails go to spam.

Is there an official list of spam words from Gmail or Yahoo? No. Neither publishes any word list. Their published requirements concern SPF, DKIM, DMARC, spam-rate ceilings, one-click unsubscribe and sending-infrastructure hygiene, which is itself the strongest evidence that words are not the lever.

Should I still avoid spammy-sounding copy? Write honestly and skip the pressure tactics, yes, but as audience craft and complaint prevention rather than filter evasion. Deceptive subjects and manufactured urgency generate the spam reports that genuinely hurt you, independent of any scoring of the words themselves.

Do spam checkers that scan my copy for bad words work? They test your message against word-era heuristics, so they can flag genuinely spam-shaped structure, but a passing grade says little about placement at Gmail or Yahoo. Testing actual placement with seed mailboxes, and your DNS with real checkers, measures what matters.

Delete the spreadsheet without guilt. Then spend the recovered attention on the checks that move placement, starting with the email authentication checker.