Search for "spam words" and you will find long lists: free, guarantee, act now, winner, cash, limited time, click here and hundreds more. The advice that follows is simple and appealing: avoid these words and your email will reach the inbox. It is also mostly wrong. A single word almost never decides where a modern email lands. Mailbox providers use machine-learned systems that weigh the sender's reputation, whether the message is authenticated, how recipients have reacted to your earlier mail, the quality of links, the structure of the message and the context of the whole text. Some words do matter, because they are signs of the deceptive or aggressive style that recipients dislike, but they matter as part of a pattern, and they matter much less than the fundamentals. This guide separates the myth from the useful part. It explains what filters actually look at, which content patterns do cause trouble, how to write persuasively without the tricks, and a checklist to run before every send.
How content filters work today
Early spam filters were rule lists. Each suspicious word added points, and a message that crossed a threshold was marked as spam. That is where the word lists come from. Modern filters combine hundreds of signals and learn from how users treat mail. They look at who the sender is and what history that domain and IP address have; whether SPF, DKIM and DMARC pass and align; how many recipients mark similar mail as spam, ignore it, delete it unread or move it to the inbox; the links and domains in the message and their reputation; the ratio of text to images; the HTML structure; the presence of a working unsubscribe mechanism; and the text itself, interpreted by language models that understand phrases in context. A message that says "free" in the sentence "free shipping on orders over $50" for a store whose customers love its mail is not treated as a message that says "FREE!!! CLICK NOW" from an unknown domain, even though both contain the word. The detailed picture is in content and spam filters.
What matters more than words
| Signal | Why it matters more than any word | Where to learn more |
|---|---|---|
| Sender reputation | Past behaviour of your domain and IP decides how much benefit of the doubt you get | Email reputation |
| Authentication | Unauthenticated mail is treated with suspicion whatever it says | Authenticating your domain |
| Engagement | Opens (with caution), clicks, replies and "not spam" actions raise placement; deletions and complaints lower it | Complaint rates |
| List quality | Old, bought or unconfirmed lists produce complaints, bounces and spam traps | List hygiene |
| Links | Shortened, mismatched or recently registered domains in links are strong negative signs | Use your own tracking domain |
| Unsubscribe | A visible, working unsubscribe reduces complaints; bulk senders also need one-click | One-click unsubscribe |
Content patterns that do cause trouble
Individual words rarely cause problems, but some styles of writing do. The common thread is that they imitate scams or pressure the reader. They fall into these groups.
- Deception. A subject line that implies a reply ("Re:", "Fwd:") when there was none; a sender name that suggests a person or company you are not; a subject line unrelated to the content. These cause complaints, and are also illegal in many places under laws such as CAN-SPAM, which prohibits deceptive headers and subject lines.
- Shouting. Subjects and bodies in all capitals, rows of exclamation marks, many currency symbols, and emphasis in every line.
- Extreme claims. Promises of guaranteed income, miracle results, risk-free offers with no terms, or prizes that need no action.
- Pressure with no substance. "Act now", "urgent" and countdown language when there is no real deadline. Genuine deadlines are fine when you explain them.
- Obfuscation. Spelling words with symbols or spaces to evade filters ("fr33", "c l i c k"), hidden text, text in the same colour as the background, and tiny fonts. Filters treat these as signs of evasion, and they hurt accessibility.
- Image-only emails. One large image with little text and no alt text. Filters cannot read the content, and recipients with images blocked see nothing.
- Sketchy links and attachments. Links through multiple redirects, link shorteners, or executable and archive attachments, especially from a new sender.
- Mismatched identity. A From domain that differs from the domain in the links and the unsubscribe URL, with no clear connection.
Notice that none of these is a word. They are behaviours. If you avoid them, you can write the word "free" when it is true.
Words that deserve care, and why
Some words are risky only in certain contexts. The table lists categories you should treat with judgement. It is not a ban list.
| Category | Examples | How to handle it |
|---|---|---|
| Financial promises | guaranteed return, earn money fast, no credit check | Avoid unless your business is regulated and you can state terms clearly |
| Health claims | cure, miracle, lose weight fast | Avoid unsupported claims; follow advertising rules |
| Prizes and giveaways | winner, you have been selected, claim your prize | Use only for real contests, with rules and a clear explanation |
| Pressure | act now, last chance, urgent | Use real deadlines with dates and reasons |
| Generic bait | click here, open immediately | Prefer descriptive link text such as "View your invoice" |
| Price words | free, cheap, discount | Fine when specific and accurate: "Free shipping over $50" |
| Security impersonation | verify your account, suspended, confirm password | Only in genuine security messages, from a clearly authenticated domain with no links to unrelated sites |
The last category matters because phishing uses it, and receivers watch it closely. A genuine "verify your account" message from your service may need extra care: authenticate fully, link only to your own domain and avoid text that looks like a bank alert. See how phishing emails look to understand what filters guard against.
Writing persuasive copy without tricks
Good copy is the best protection. State who you are and why the person is receiving the message in the first lines. Write a subject line that says what is inside, in normal sentence case, and keep the preheader in sync with it; see writing subject lines. Use one main call to action and descriptive link text. Keep the balance of text and images sensible, give every image alt text and include a plain-text version. Put a real postal or contact identity and an unsubscribe link in the footer; see email footer examples. Personalise with data you really have, and provide fallbacks for missing fields: personalisation variables and fallbacks. If the offer is good and the audience asked for it, the words matter very little.
A pre-send checklist
- Is every recipient someone who asked for this kind of email, and recently?
- Do SPF, DKIM and DMARC pass and align for the From domain?
- Is the sender name recognisable, and does the subject match the content?
- Is there a plain-text part, alt text for images and a sensible text-to-image balance?
- Do all links go to domains you own or trust, with no shorteners, and does the link text describe the destination?
- Is the unsubscribe link visible and working, and does the message include the list-unsubscribe header for bulk mail?
- Have you sent a test to seed addresses at the major providers and checked where it landed? See inbox placement testing.
- Have you checked the content in a spam checker, as one input, not as the verdict? See spam checkers.
- Are you sending to your most engaged segment first, with volume that suits your history?
When mail still goes to spam
If you follow all of the above and your messages still land in spam, the cause is almost never a single word. Look at reputation, list quality, authentication, complaints and blocklists. Work through the spam troubleshooting checklist and, for rejections with an error code, SMTP error 554. Compare the campaigns that landed well with those that did not: differences in audience and engagement usually explain more than differences in copy.
Send to people who want your mail Delivery logs, complaint tracking and authentication tools help you keep content and reputation healthy.
References and further reading
Frequently asked questions
Do spam trigger words really send email to spam?
Rarely on their own. Modern filters weigh reputation, authentication, engagement, links and the whole message. Deceptive or pressuring patterns matter more than any single word.
Can I use the word free in an email?
Yes, when it is true and specific, such as free shipping over a stated amount. The problem is deceptive or exaggerated use, not the word.
What content actually hurts deliverability?
Deceptive subjects and sender names, shouting, extreme claims, hidden text, image-only messages, shortened or mismatched links and missing unsubscribe options.
How do I test whether my email will go to spam?
Authenticate your domain, send to seed addresses at major providers, use a spam checker as one input and monitor engagement and complaints.
Do emojis or exclamation marks cause spam filtering?
A single emoji or exclamation mark does not. Excess and patterns that look like shouting or pressure can add to the risk.