Diagnose · belegte Antwort
Spam Terms in Email: Which Words Filters Really Score
TL;DR: Spam terms are phrases common in spam, like 'Dear Friend', '100% guaranteed', 'unclaimed funds', and 'act now'. In SpamAssassin, the only filter with public weights, a few phrases cost 2 to 3 points against a 5.0 threshold while most cost under 0.6. Gmail and Outlook rely mainly on authentication, reputation, and complaints, so fix those first, then remove deceptive, high-weight phrases.
What are spam trigger words?
Spam trigger words are words and phrases that appear far more often in spam than in wanted email, such as 'Dear Friend', '100% guaranteed', 'unclaimed funds', 'act now', and 'free instant quote'. Content filters learn or encode these patterns and add points when they appear. Lists of them circulate widely: ActiveCampaign publishes 188, Vero 100, and Pipedrive, Mailjet, Benchmark, and Mailtrap their own. Most lists are compiled by marketers, not filter vendors, and none publishes how much each word counts. The one filter whose rules and weights are public is Apache SpamAssassin, used by many business, university, and hosting mail servers, so it is the best evidence of how phrases are actually scored. Its rule set shows something the lists skip: a few phrases are expensive, most cost almost nothing, and no single word sends a message to spam on its own. Gmail, Outlook.com, and Yahoo use machine-learning filters that weigh sender reputation, authentication, and recipient behavior far more than individual words.
Which spam words does SpamAssassin actually score?
SpamAssassin scores specific phrase patterns, and the weights below come from Apache's current rule set using score set 3, the normal production configuration with Bayes and network tests on. The default spam threshold is 5.0 points. Salutations and scam language are the costliest: 'Dear Friend' at the start of a line (DEAR_FRIEND) adds 2.577, '100% guaranteed' (GUARANTEED_100_PERCENT) 2.699, and 'unclaimed' followed by funds, money, prizes, or assets (UNCLAIMED_MONEY) 2.427. Advance-fee patterns about large sums add 2 to 3 points, and 'Dear sir', 'Dear madam', or 'Dear investor' (DEAR_SOMETHING) adds 1.973. Commercial phrases are much cheaper: 'free instant quote' or 'free no-obligation quote' adds 1.297, a subject starting with 'Buy' adds 0.639, 'urgent matter' 0.573, an all-caps subject 0.5, and 'ACT NOW' in capitals only 0.1. A body that is 75 to 100% uppercase costs 1.48 without Bayes but 0.001 with it, because the trained classifier makes the crude rule unnecessary.
Rule Score (set 3) Pattern it matches
GUARANTEED_100_PERCENT 2.699 "100% guaranteed"
DEAR_FRIEND 2.577 line starting "Dear Friend"
UNCLAIMED_MONEY 2.427 "unclaimed funds / money / prizes / assets"
ADVANCE_FEE_3_NEW_MONEY 3.099 several advance-fee money phrases together
MILLION_USD 1.999 "million dollars" style amounts
DRUGS_ERECTILE 1.994 erectile drug names
DEAR_SOMETHING 1.973 "Dear sir / madam / investor / candidate"
FREE_QUOTE_INSTANT 1.297 "free instant / express / no-obligation quote"
ONLINE_PHARMACY 0.650 "online pharmacy"
SUBJ_BUY 0.639 subject starts with "Buy"
URG_BIZ 0.573 "urgent matter"
SUBJ_ALL_CAPS 0.5 subject in all capitals
ACT_NOW_CAPS 0.1 "ACT NOW" in capitals
PLING_QUERY 0.1 subject with both ! and ?
LOTS_OF_MONEY 0.010 large money amountsWhat spam words should you avoid in email?
Avoid spam words in groups by what they signal, not one by one. Scam and fraud signals carry the highest scores and also cost trust: 'Dear Friend', 'Dear Sir/Madam', 'unclaimed funds', 'beneficiary', 'wire transfer', 'inheritance', 'lottery', and 'you have been selected'. Exaggerated claims: '100% guaranteed', 'risk-free', 'no catch', 'miracle', 'once in a lifetime', and 'double your income'. Pressure and urgency: 'act now', 'urgent', 'limited time only', 'expires today', 'don't delete', and 'final notice' when it is not final. Money and price bait: 'free money', 'cash bonus', 'earn $', 'lowest price', 'no fees', and 'pre-approved'. Pharmacy and adult terms: drug names, 'online pharmacy', and 'weight loss' miracle language. Formatting signals: ALL-CAPS subjects, multiple exclamation marks, '!?' combinations, and fake 'Re:' or 'Fwd:' prefixes. The common theme is deception, which is also what the law and Gmail target, so honest copy rarely needs these phrases at all.
Do spam words really send email to spam?
Spam words alone rarely send email to spam at large mailbox providers. Gmail, Outlook.com, and Yahoo classify mail mainly on sender reputation, SPF, DKIM, and DMARC authentication, complaint rates, and how recipients interact with past messages; a domain with a clean history and a spam rate under Google's 0.30% ceiling can send 'Free shipping, limited time only' to the inbox every week. Where words matter is at the margin and on rule-based filters. On a SpamAssassin server with the default 5.0 threshold, a message with infrastructure problems, such as no reverse DNS (+0.793), a missing Date header (+1.360), and a link to a blocklisted domain (+2.5), is already at 4.65 before any copy is read, and a single 'Dear Friend' then tips it over. For a well-authenticated message from a reputable sender, the same phrase leaves it near 2.5, comfortably below. So fix authentication and list quality first, then remove high-weight phrases; word-swapping on a poorly authenticated domain does not help.
What do Gmail's rules and the law say about deceptive wording?
Gmail's email sender guidelines and US law both target deceptive wording rather than specific words. Google says message headers and content should be accurate and not misleading; subject lines should not start with 'Re:' or 'Fwd:' unless the message is an actual reply or forward; display names should not contain subject or message content; and you should not use HTML or CSS to hide content, which might cause messages to be marked as spam. Under the US CAN-SPAM Act, the FTC prohibits deceptive subject lines that do not reflect the message content, requires a valid physical postal address and a working opt-out honored within 10 business days, and sets penalties of up to $53,088 per violating email. That is why the phrases with the highest SpamAssassin weights, fake urgency, fake guarantees, and fake familiarity, are the same ones that create legal and reputation risk. A subject line like 'Re: your invoice' on a promotion fails all three at once.
How do you test a message before sending?
Check an email for spam words by scoring the full message rather than pasting copy into a word list, because filters score the subject, body, headers, links, and HTML together. Three methods work. First, run SpamAssassin locally: save the delivered message with full headers as an .eml file and run spamassassin -t, which lists each rule that fired, including the phrase rules above, with its points. Second, use a seed-based tester such as mail-tester.com or MailerCheck, which shows SpamAssassin results with SPF, DKIM, DMARC, and blocklist checks. Third, for Gmail specifically, watch the spam rate in Google Postmaster Tools after sending: no tester predicts Gmail's machine-learning filter, but a rising user-reported spam rate shows that recipients, not words, are the problem. Online 'spam word checkers' that highlight words without scores are useful only as a proofreading pass. Whatever the tool, change what scores highest first and retest.
# Score a real message with the current Apache rule set
sudo sa-update
spamassassin -t < campaign.eml | grep -E '^ *-?[0-9.]+ [A-Z0-9_]+'
# 2.6 DEAR_FRIEND BODY: Dear Friend? That's not very dear!
# 0.5 SUBJ_ALL_CAPS Subject is all capitals
# -0.1 DKIM_VALID Message has at least one valid DKIM signatureWhat should you write instead of spammy phrases?
Write specific, accurate copy instead of spammy phrases: concrete details read as legitimate to both filters and people. Replace 'Dear Friend' or 'Dear Sir/Madam' with the recipient's name or no salutation. Replace '100% guaranteed' with the actual policy: '30-day refund, no questions asked'. Replace 'Act now!' or 'Urgent' with the real deadline: 'Pricing changes on November 1'. Replace 'Free instant quote' with what happens next: 'Get a quote in about 2 minutes'. Replace 'Buy now' at the start of a subject with the benefit or the product name. Use sentence case instead of capitals and one exclamation mark at most. For transactional email, where words are rarely the issue, lead with the event ('Your password was changed', 'Invoice INV-1042 is ready') and keep promotions out of the message; mixing marketing into receipts and resets is a bigger risk than any single word, because it turns a message recipients expect into one they may report.
Häufige Fragen von Teams
Is the word 'free' a spam word?
Not by itself. SpamAssassin's current rules score phrases like 'free instant quote' (+1.297 in score set 3), not the word 'free' alone, and Gmail does not publish word penalties. 'Free shipping' in an honest, authenticated email from a known sender is fine.
How many spam words are too many?
There is no fixed count. Filters add up weighted signals, so one high-weight phrase such as 'Dear Friend' can matter more than several low-weight ones. Score the full message with SpamAssassin and keep the total well below 5.0, ideally under 2.
Do spam words matter in transactional emails?
Less than in marketing email, because receipts and password resets use plain, factual language. The bigger risks are missing authentication, broken HTML, and adding promotions to transactional messages. Keep them short, specific, and free of marketing content.
Are spam words in the subject line worse than in the body?
Some rules apply only to the subject, such as an all-caps subject (+0.5), a subject starting with 'Buy' (+0.639), or '!?' (+0.1). Gmail also checks that subjects are not misleading, such as fake 'Re:' prefixes, so subjects deserve the most care.
Can I use spam words if I have good sender reputation?
Reputation lowers the risk at Gmail, Outlook, and Yahoo, but business servers running SpamAssassin still add the phrase points. Deceptive phrases also raise complaint rates and CAN-SPAM risk, so drop them regardless of reputation.
Primärquellen
- SpamAssassin body test rules — Apache SpamAssassin
- SpamAssassin generated scores — Apache SpamAssassin
- Email sender guidelines — Google
- CAN-SPAM Act: A Compliance Guide for Business — US Federal Trade Commission
- 188 spam words to avoid — ActiveCampaign