Content Psychology & Filtering Signals

Many senders believe certain words send email to spam. At the big mailbox providers that is mostly a myth. Your reputation and your links matter far more than your wording. This article shows which parts of your content filters really look at, and how your content shapes the way readers react, which matters even more.

The “spam words” myth

You may have heard that the word “free”, a subject in ALL CAPS or too many images will send your email to junk. These ideas come from how spam filters worked around 20 years ago. Older filters, like early versions of SpamAssassin, gave each message points for phrases such as “free”, “buy now” or “act fast”. Enough points, and the message went to spam.

Gmail, Outlook and Yahoo now use machine learning models trained on huge numbers of messages. These models look at the whole picture: who sent the message, how it is built, what it links to and how readers react. A single word rarely decides anything.

Some rule-based filters still exist, especially on company mail servers and smaller providers. Even there, words are only one small part of the score.

Content still matters. It just matters less than most senders think, and in different ways.

What filters check in your content

Filters look at how your email is built, not at single words.

HTML quality

HTML is the code that lays out your email.

• Very messy HTML, such as deeply nested tables or broken tags, can count slightly against you.
• HTML that looks like templates used in known spam campaigns gets flagged.
• Clean, simple HTML is neutral. It won’t help, but it won’t hurt either.

Text and images

• An email that is only an image, with no real text, looks suspicious. Spammers use images to hide words from filters.
• A sensible mix of text and images is fine.
• There is no proven “golden ratio” of text to images.

Links

Links are one of the strongest content signals.

• Filters check every link against reputation lists such as Google Safe Browsing, URIBL and SURBL. These lists track domains used for spam, phishing or malware.
• One link to a listed domain can sink an otherwise clean email.
• Link shorteners like bit.ly hide where a link goes. Spammers use them heavily, so filters distrust them. Link to your own domain instead.
• Long chains of tracking redirects (links that bounce through several addresses before arriving) are harder to check and can look suspicious.

Attachments

• Program files such as .exe, .bat or .scr are almost always blocked.
• Password-protected zip files get flagged, because filters can’t look inside for malware.
• PDFs and documents are scanned for harmful content.
• Large files aren’t a spam signal by themselves, but they slow delivery. For big lists, link to the file instead of attaching it.

Content that doesn’t fit your history

Filters compare your email with what you usually send.

• A sudden change. If a trusted sender suddenly sends something quite different, such as a new template or new link domains, filters may look more closely.
• Looks like recent spam. If your email’s layout, links or headers closely match spam that was just flagged, the model rates it as riskier. This is called fingerprinting, and it compares structure, not words.
• Past reactions. Providers can learn from how readers reacted to similar emails from you before. If similar emails were mostly ignored, the next one may start at a disadvantage.

What barely matters: single words

• “Free” in a subject line doesn’t trigger big-provider filters.
• “Act now”, “limited time” and “exclusive offer” aren’t spam triggers on their own.
• Prices, percentages and sales language are neutral.
• ALL CAPS subjects can feel pushy, so readers may react badly. That affects engagement, not the filter.
• Emoji in subjects are neutral for filtering. Some audiences like them and some don’t.

These myths survive because they were true once. Marketers saw real results when testing older filters. Those filters have since been replaced.

How content shapes reader behaviour

Content rarely sends an email to spam directly. But it shapes how readers react, and providers watch those reactions closely. Opens, clicks, replies and “Not spam” rescues count in your favour. Deletes without reading and spam reports count against you.

Subject lines

The subject line decides whether someone opens your email.

• Clear beats clever. “Your March account summary” works better than “You won’t believe what happened this month”.
• Make it personal in a useful way. Adding someone’s first name is weak. “The jacket you saved is back in stock” is strong, because it shows the email is about them.
• Only use real urgency. “Your subscription ends tomorrow” helps. A fake “LAST CHANCE” erodes trust and brings more complaints over time.
• Put the key words first. Phones cut off long subject lines, so the start needs to carry the message.
• Write the preview text. The preview text (or preheader) is the line that shows after the subject in the inbox. Many senders leave it as leftover code or “View in browser”. Use it to say what’s inside.

The From name

People often decide whether to open by looking at who sent the email, before they read the subject.

• Use a From name people recognise, and include your brand.
• Keep it the same from email to email.
• Changing it often confuses people and can lead to more complaints.

The body of the email

The body decides what happens after the open: a click, a reply or a quick close.

• Show the value early. People scan rather than read. Say why the email matters in the first two or three sentences.
• Ask for one thing. One clear call to action, like a single “Track your order” button, works better than several competing ones.
• Make it easy to scan. Use short paragraphs, bullet points and a few bold phrases.
• Design for phones. Many readers open email on a phone. Avoid tiny text, sideways scrolling and anything that needs a mouse hover.

Content that does hurt

Single words don’t matter much, but some content choices really do raise your spam risk.

Misleading content

• Fake reply subjects. Starting a subject with “RE:” when there was no earlier conversation is deceptive. Providers increasingly spot it.
• False urgency. “Only 3 left!” when you have thousands teaches readers to distrust you. Engagement falls over time.
• Phishing signs. Phishing is email that tricks people into giving away passwords or money. Emails that copy login pages, ask for passwords or imitate alerts from other companies trigger special phishing filters.

Technical problems

• Broken tracking links. Some providers follow links to check them. If your redirects fail or time out, the email gets flagged.
• Hidden text. White text on a white background, or text with a font size of zero, is an old spam trick. Filters catch it easily.
• Too many tracking pixels. A tracking pixel is a tiny invisible image that reports when the email is opened. One per email is normal. Several from different services add clutter.
• JavaScript. Most email apps strip it out. Including it achieves nothing and can trigger security filters.

Risky patterns

• Affiliate links. Affiliate networks that spammers also use have poor link reputation. If you use affiliate links, send them through a redirect on your own domain.
• Dozens of links. An email packed with links looks unusual and draws a closer look.
• Attachments to large lists. Sending big files to many people is unusual. Link to hosted files instead.

Testing your content

Content effects are small and depend on your audience. Testing is the only way to know what works for yours.

A/B tests

An A/B test sends two versions to random halves of a group and compares the results.

Subject lines

• Change one thing at a time: length, personalisation, urgency, or a question versus a statement.
• Use large groups, often 1,000 people or more per version, so the result isn’t luck.
• Run each test over the same time window before picking a winner.
• Treat open rates with care. Apple Mail’s privacy feature loads images automatically, which counts many emails as “opened” whether or not anyone read them. Clicks and replies are more reliable.

The email body

• Test layouts, such as one column versus two, or image-heavy versus text-heavy.
• Test where your button goes and what it says.
• Measure clicks, and what people do after the click, such as buying.

How often you send

• Split your list into random groups and send to each at a different frequency.
• Run the test for at least four weeks. Fatigue takes time to show.
• Compare unsubscribes and complaints for each group, not only clicks.

Check inbox placement after changes

After a big content change, check where your email lands:

• Send to test inboxes at Gmail, Outlook, Yahoo and other providers you care about. Mailwel’s test inbox checks one message in detail.
• Compare placement before and after the change.
• Check each provider separately. They don’t all react the same way.

What matters most

A rough order of what matters for content-related deliverability, from most to least:

1. Sender reputation. It outweighs every content signal.
2. Link reputation. The domains your links point to.
3. Structure. How closely your email resembles known spam templates.
4. HTML quality. Broken, overly complex or suspicious code.
5. Reader reactions. How people respond to the content.
6. An honest subject. Does it match the body?
7. Visual tricks. Image-only emails, hidden text, misleading layouts.
8. Specific words. Almost irrelevant at the big providers.

Checklist

• Write content your readers find useful, and ask for one clear action.
• Use a recognisable From name and an honest subject line.
• Keep your HTML clean and include real text, not just images.
• Link only to domains you trust. Avoid link shorteners.
• Never hide text or pretend to be someone else.
• Retest inbox placement after big template or content changes.

Stop worrying about “spam words”. Focus on reputation, reader engagement and a clean list.