Spam trigger words in cold email
Word lists are a distraction from the real filters. Receiving systems weigh authentication, sender history, recipient behaviour and message structure far more than any single phrase. Some language still matters, but it matters as a signal of bulk mail, not as a magic list to avoid.
Updated October 26, 2026
Why word lists mislead
Every published list of trigger words contradicts another one, because the words have no universal weight. The same phrase in a one-to-one message from a known sender passes; in a thousand-recipient blast from a new domain it does not. The variable is the context, not the word.
Chasing a word list produces unnatural copy, which is itself a signal. A message written to avoid imaginary filters reads like it was assembled by a machine, and recipients respond accordingly.
- Filters score combinations and sender history, not isolated words.
- Unnatural copy reduces replies, which is the outcome that matters.
- A phrase is risky when it promises money, urgency or certainty.
The stronger inputs are covered in sender reputation, and they are the ones worth fixing first.
The language that genuinely causes trouble
Four patterns recur: promises of income, manufactured urgency, absolute guarantees, and the vocabulary of bulk marketing. They are risky because they describe the message as an advertisement, and because recipients who did not expect it mark it as one.
The fix is not a synonym. Replace the claim with something specific and true. Instead of a guarantee, describe what the product does. Instead of urgency, give a reason the timing is relevant to the recipient.
- Income claims and get-rich framing.
- Countdowns, expiring offers and false deadlines.
- Guarantees of results, rankings or delivery.
- Bulk vocabulary: newsletter, promo, limited time, act now.
Structure matters more than vocabulary
A plain message with one link and no tracking is structurally different from an HTML template with an image header, five links and a hidden pixel. The first looks like correspondence; the second looks like a campaign. If a message is being filtered, change its shape before you change its adjectives.
Attachments are a separate risk. A first-contact email with a PDF is a common malware shape, and some gateways will quarantine it regardless of content. Link to a page instead, or send nothing and let the reply start the conversation.
Shape, authentication and list quality are one procedure: deliverability checklist.
Rewriting a flagged message without gutting it
Start from the sentence a recipient would actually answer. “Would a 15-minute call about your Q3 pipeline be useful?” is specific, answerable and contains no trigger vocabulary. The promotional version of the same idea usually contains three claims and no question.
Worked example: a message that opened with an income promise and closed with a deadline was rewritten to name the recipient's industry, describe one concrete outcome the sender had seen, and ask a single question. The rewrite removed every trigger phrase, kept the offer, and produced replies where the original had none.
- Keep one idea and one question.
- Name the recipient's context, not your product's category.
- Remove every deadline you cannot justify.
- Read it aloud; if it sounds like an advert, rewrite it.
If you want the sending handled while you fix the copy, cold email campaigns runs campaigns on managed mailboxes.
When a message is filtered anyway, in what order to look
Start with authentication, because everything else is wasted effort if SPF, DKIM or DMARC are misaligned. Next, the list: a verified, recently checked audience removes the largest variable. Then the shape of the message, since a template with an image header and several links behaves differently from plain text. Only then read the copy.
This order exists because copy is the easiest thing to change and the least likely cause. Rewriting a subject three times while the domain fails alignment produces three failures and no information.
If the message still filters after those steps, test one variable at a time with a small verified segment. Changing the subject, the body and the link together tells you nothing, even when the next batch lands in the inbox.
- Authentication, list, shape, then copy.
- One change per test batch.
- Keep the winning version until a few hundred sends confirm it.
- Terms that describe a commercial relationship are not automatically risky: pricing, invoice and contract are ordinary words in business mail. The risk comes from claiming an outcome you cannot prove, not from naming a document.
- A better use of time than a word list is a checklist of what the message must contain: a true statement about the recipient, one reason to answer, one question, and a sender identity a person could reply to. If a draft lacks those, no amount of word substitution will fix it. Keep the checklist beside the composer and read it before every batch, because the failure mode is usually a rushed message rather than a forbidden adjective.
Risky pattern, why it is risky and a truthful replacement
| Pattern | Why it is risky | Replacement |
|---|---|---|
| Guaranteed results | Unverifiable claim, filters and readers penalise it | Describe what the product does |
| Act now / limited time | Manufactured urgency | State a real reason the timing matters |
| Make $X per month | Income claim | Describe the work, not the payout |
| Free audit / free gift | Bait framing | Offer a specific, small next step |
| HTML banner + five links | Bulk-message shape | Plain text, one link or none |
The right-hand column is not a trick to evade filters; it is the version you can defend if a recipient asks.
What rewriting will not fix
Copy is one input among several.
- It will not repair missing authentication or a blocklisted domain.
- It will not make an unverified or non-consenting list safe.
- It will not guarantee inbox placement; that decision belongs to the receiver.
- It will not compensate for sending too fast from a new mailbox.
Frequently asked questions
Is there a definitive list of spam words?
No. Lists conflict because filters weigh context, sender history and structure. Treat any word list as a rough writing prompt, not a rulebook.
Should I avoid the word free entirely?
No. "Free trial" is ordinary language. It becomes risky when it is paired with urgency and a claim you cannot support.
Do links cause spam filtering?
A single ordinary link is normal. Redirect chains, shortened URLs and many links in a first-contact message are the patterns that draw attention.
Does plain text always deliver better?
It is simpler and closer to personal mail, but it is one factor. Authentication, list quality and pacing still decide most of the outcome.
Ready to run a campaign? Compare published plans or create a free account. Trials cannot send until a paid plan is active.