Content and design: what actually matters

Content affects deliverability less than the folklore says and more than zero. That's the whole verdict, so let me size both halves precisely. Modern filters weigh who is sending (reputation, covered in the sender reputation lesson) and whether the identity checks out (authentication) far more heavily than what the message says. The same words deliver perfectly from a trusted sender and folder from a distrusted one, which tells you content can't be the main term in the equation. But filters do read the message, and a handful of structural choices hand them easy reasons to doubt you. This lesson is about those choices.

The "spam words" lists are folklore

You've seen the articles: 200 words to never use, and a promise that "free" in a subject line sends you to spam. Treat these lists as superstition. Twenty years ago, filters really did lean on word scoring, and spammers really did write "F R E E" with spaces to dodge it. That arms race is over. Today a bank writes "free" in every statement notification and delivers fine, and receipts say "order confirmation" without incident from trusted senders everywhere.

The myth survives because content is the one factor the sender can see. You can't inspect your reputation directly, but you can stare at your subject line, so that's where the blame lands. It's the drunk looking for keys under the streetlight.

The honest concession: filters absolutely read content, as one signal among many. A message whose every structural feature matches known spam patterns starts the race with a handicap. The difference from the word-list myth is that the patterns that matter are structural (images, links, formats, the shapes described below), while individual vocabulary on an otherwise clean message from a reputable sender moves almost nothing. Write for humans. Filter-driven synonym swapping ("f.r.e.e") reads as evasion, and evasion is itself a spam pattern.

The one-giant-image email

The classic content mistake is the email that is a single large image with little or no real text. Designers love it because it renders pixel-perfect. Filters distrust it for a specific historical reason: image-only mail was a spam evasion technique that hid the pitch inside pixels where text analysis couldn't read it. Mail with a very high image-to-text ratio still matches that shape.

It's also just bad email. Images are blocked by default in some clients, so those readers see a blank rectangle. Screen readers have nothing to read. And your one shot at engagement rides on a slow-loading image over hotel wifi.

You don't need a magic ratio, and anyone selling you an exact percentage is guessing. The working rule: real text carries the message, images illustrate it. If the message survives with images turned off, you're fine.

Link hygiene

Filters look hard at links, because links are where the harm in malicious mail lives. Three patterns to avoid.

Link shorteners first. A bit.ly or similar shortened URL hides its destination, which is precisely why phishing mail loves shorteners, and a shortener's domain carries the pooled reputation of everyone who uses it, spammers included. In bulk mail there is no legitimate need: your ESP already wraps links in its own tracking domain. Shorteners in a campaign are all downside.

Second, link domains that don't match the sender. Mail from yourbrand.com where every link points somewhere unrelated has the shape of phishing, since impersonation mail by definition sends from one identity and links to another. Ideally your ESP's click-tracking domain is a branded subdomain of your own (most ESPs support this), so sender and links agree.

Third, too many distinct domains. A newsletter linking out to a handful of destinations is normal. A message sprinkling links across a dozen unrelated domains inherits the reputation of every one of them, and it only takes one blocklisted domain in the pile. This is worth an internal audit occasionally: affiliate links and partner links accumulate alongside old footer links, and nobody rechecks where they point.

Every email you send to your Inbox Decay test address gets a free content snapshot that covers exactly this ground: link count, where the links point, whether shorteners are involved, and the link domains checked against domain blocklists. The full list of what the snapshot measures is in every check, explained. The blocklisted-link-domain case is a quietly nasty one, because the domain you linked to can get listed months after you added the link to your template, and nothing on your side changed.

Attachments don't belong in bulk mail

Attachments are the main vehicle for malware, so filters treat them with suspicion in direct proportion to how unusual they are. A person sending a person a PDF is normal. A bulk campaign carrying an attachment to thousands of recipients is a shape filters rarely see from legitimate senders, because legitimate senders host the file and link to it.

So do that. Host the PDF, link the PDF. The message gets lighter, and you stop resembling a malware campaign. This one is close to an absolute rule, which is rare in deliverability.

The unsubscribe must be visible

There are two unsubscribe mechanisms and you need both. The List-Unsubscribe header (with its one-click variant) lives in the message's technical headers and powers the unsubscribe button mailbox providers render in their own interface; the 2024 bulk sender rules require it, and the bulk sender rules lesson covers it. This lesson is about the other one: a visible unsubscribe link in the body of the message.

The logic is brutally simple. A reader who wants out will find an exit. If the unsubscribe link is present and findable, that exit costs you one subscriber. If it's missing or hidden in grey-on-grey four-point type, the exit they find instead is the report-spam button, and that costs reputation on every future send. An unsubscribe from someone who stopped wanting your mail is a good outcome. Burying the link doesn't retain a reader; it converts a quiet goodbye into a complaint.

The plain-text part

A well-formed marketing email is two messages in one envelope: an HTML version and a plain-text version of the same content, bundled in a MIME structure called multipart/alternative (the format from anatomy of an email). The recipient's client picks the one it can render.

HTML-only mail, with no plain-text part, is a small negative signal with an easy fix. Every serious sending platform generates the text part automatically, so its absence suggests hand-rolled sending software of the kind spammers write in a hurry. Check that your ESP does this (nearly all do) and, if you ever build sending yourself, include it. Cheap insurance, and it's also what some watches, terminal clients and screen readers actually display.

Preview text

Preview text (also called the preheader) is the snippet an inbox shows after the subject line. Filters don't score it in any way worth planning around. Humans do, and humans are what filters measure.

An inbox listing gives you two lines to earn an open: subject, then preview. Leave the preview to chance and it fills with the first text in your message, which is typically "View this email in your browser" or raw menu links. That's a wasted line at the exact moment a reader decides between opening and deleting unread, and every unread delete teaches the filter a little about how wanted your mail is. Set the preview text deliberately, like a second subject line. This is the cleanest example of how content really affects deliverability in 2026: through human behavior, one step removed, rather than through word-scoring.

How to tell content is your problem

Content problems are specific: one campaign folders while its neighbors deliver. That pattern, which we see regularly when people bring us a "sudden" spam problem, points at something in that send: the giant image, the shortener, the sketchy link domain, the attachment. When everything you send folders, stop auditing your copy, because the cause is upstream in reputation, authentication, list quality or infrastructure. Why emails suddenly go to spam walks that diagnosis in order.

Content is the polish, in other words. Worth an afternoon of template fixes, and never a substitute for the foundations underneath it.

Even a flawless message inherits the standing of every other kind of mail sent under the same name, which is how a marketing blast ends up taxing the password resets that must arrive. Containing that damage is the work of separating your mail streams.

Terms from this lesson

• image-to-text ratio - the balance of image area to real text in a message. Very image-heavy mail matches a historical spam evasion pattern and draws suspicion.
• link shortener - a service that turns a URL into a short redirect on the service's own domain, so the destination is hidden. Pooled reputation and hidden targets make them a negative in bulk mail.
• link domain - the domain a link in your message points to. Filters evaluate the reputation of every domain you link, and a blocklisted one taints the message.
• plain-text part - the text-only copy of a message bundled alongside the HTML version inside multipart/alternative. Its absence is a small spammer-shaped signal.
• preview text (preheader) - the snippet shown after the subject line in the inbox listing. Invisible to filters as such, decisive for the human open-or-delete choice they measure.
• visible unsubscribe - an easy-to-find opt-out link in the message body. Its absence converts quiet goodbyes into report-spam complaints.
• content snapshot - the free, programmatic read Inbox Decay runs on each email sent to your test address: structure, image balance, links and their domains, attachments, unsubscribe visibility.

Check yourself

1. How much do "spam word" lists matter to modern filters?

A single banned word sends a message to spamThey matter only for senders using dedicated IPsFilters no longer read message content at allAlmost nothing on a clean message from a reputable sender; structural patterns and reputation carry the weight

2. Why do filters distrust an email that is one large image with little text?

Image-only mail was a technique for hiding the pitch from text analysis, and the shape still matchesImages consume too much server bandwidthLarge images always fail DKIM signingThe 2024 bulk sender rules ban images

3. What is wrong with using a link shortener in a bulk campaign?

Shortened links cannot be clicked in most email clientsIt breaks the List-Unsubscribe headerIt hides the destination and carries the pooled reputation of every spammer using the same shortenerIt increases the message size

4. Your last campaign went to spam but the five before it delivered fine. What does that pattern suggest?

Your domain reputation has collapsedSomething in that specific message, in its images, links, formats or attachments, gave filters a reasonYour SPF record brokeYour list has aged past usefulness