The spam-word list is the most durable myth in cold email. Modern filters are statistical and reputation-weighted; they are not scanning for “free” and adding points. A sender with good reputation can write “guarantee” and land in the inbox, and a sender with bad reputation can write beautifully and land in spam.
That said, copy is not irrelevant, it just matters in a narrower and more specific way than the lists suggest.
What actually matters in the message
| Element | Effect | Rule |
|---|---|---|
| Number of links | Real | One, or none in a first email. Several links is the single strongest commercial signal |
| Images | Real | None in cold email. An image-heavy message reads as marketing to filters and to people |
| Tracking pixels | Real | Consider turning open tracking off entirely. It adds a pixel and buys you a metric you should not manage against |
| HTML formatting | Real | Plain text or near-plain. Coloured buttons and tables belong in newsletters |
| Attachments | Real | Never in a first email |
| Unsubscribe wording | Real, and required | Give a plain way out. Its absence is the actual violation |
| Individual “spam words” | Overstated | Do not contort your writing. Avoid the obvious financial-scam register and move on |
| ALL CAPS and multiple !!! | Mildly real | Poor writing anyway |
The one kind of copy that reliably hurts
Dishonest framing. “Following up on our conversation” when there was no conversation, “Re:” prefixes on a first email, or a fake thread history, these earn complaints, and complaints are the copy problem that actually damages a domain. Filters do not detect the lie; recipients do, and their complaint is what the filter learns from.
That is the mechanism worth internalising: copy affects placement mostly through recipient behaviour, not through word scanning. A message that provokes complaints hurts you. A message that gets replies helps you, and replies are the strongest positive signal there is.
What matters much more than copy
In rough order of impact: list quality and bounce rate, authentication, volume pacing per mailbox, domain and IP reputation, and only then the message. A campaign in spam with clean records and a verified list is a volume or reputation problem far more often than a wording problem. The causes in the order they occur.
Which is why the useful sequence when something goes wrong is: check the four records, verify the list, check volume against 25 to 30 per mailbox per day, then run a placement test. Rewriting copy first is the most common wasted week in outbound.
A short pre-send check
- One link maximum, and it resolves.
- No images, no attachments, no tracking pixel if you can live without opens.
- Plain signature, three lines of text, not a designed block.
- Nothing that implies a relationship you do not have.
- A plain sentence offering an exit.
- Read it aloud. If it sounds like marketing, it will be treated as marketing.
Next: how to write the thing in the first place, or why cold emails go to spam.
Frequently asked questions
Do spam trigger words really matter?
Far less than the lists suggest. Modern filters are statistical and reputation-weighted rather than scanning for keywords. Avoid the obvious financial-scam register, but do not contort your writing.
What in my copy actually affects placement?
Number of links, images, attachments, heavy HTML formatting, tracking pixels and whether you offer a plain way out. One link and no images is the right shape for a cold first email.
What is the worst copy mistake?
Dishonest framing, a fake “Re:” prefix, or implying a conversation that never happened. It earns complaints, and complaints are what filters learn from.
Should I turn off open tracking?
Often yes. It adds a pixel and buys a metric you should not manage against. Reply rate is the number that matters.
My campaign went to spam. Should I rewrite the copy?
Not first. Check authentication, verify the list, confirm volume is within 25 to 30 sends per mailbox per day, then run a placement test. Rewriting copy first is the most common wasted week in outbound.