Why Your Clinic Emails Land in Promotions (and the 6 Signals Causing It)

On April 15 we sent one email to 244 wellness providers. This was not a cold list. These were people who had been through training with us, people who knew our names and had our phone number, many of whom had already asked about the exact product the email was announcing. The offer was a free wholesale account with no minimums.

Zero replies.

Not a weak reply rate. Zero. When we checked the copies we could actually see, the message was sitting in the Gmail Promotions tab, filed alongside retail coupons.

The copy was not the problem. The list was not the problem. The offer was not the problem. The container was wrong, and the container decided whether anyone ever read the thing. Here is the post-mortem: what Google actually publishes about this, the six signals we found when we took the email apart, and what we rebuilt.

What Google Actually Tells You

Start with the part most marketing advice skips. Google publishes almost nothing about how the inbox tabs work.

Gmail's own help documentation defines the Promotions category in six words: "Deals, offers, and other promotional emails." It states that sorting happens automatically. It does not say how. Google's developer documentation for Promotions tab annotations is similarly quiet on classification. It explains how to decorate a message that is already in Promotions, and it warns that "A variety of factors affect whether email annotations are shown to users, including quality filters and frequency limits." Nothing about how a message gets sorted there in the first place.

What Google does publish in detail is its email sender requirements. Every sender needs SPF or DKIM authentication, valid forward and reverse DNS records, and a TLS connection, and must keep spam complaint rates in Postmaster Tools below 0.30%. Senders pushing 5,000 or more messages a day to personal Gmail accounts also need DMARC, domain alignment between the From header and the SPF or DKIM domain, and one-click unsubscribe using the List-Unsubscribe-Post and List-Unsubscribe headers.

Here is the part clinic owners miss. Those published requirements are the bar for getting delivered at all. They are not the bar for landing in the Primary tab. You can satisfy every documented requirement and still get filed under deals and offers, because Primary placement is a separate judgment about whether the message looks like a person writing to a person.

So what follows is not a documented algorithm and no one should sell it to you as one. It is a diagnosis of one real email: what it contained, what we removed, and why. Placement results vary by sending domain, list history, and individual recipient behavior.

Signal 1: The Email Was Built Out of Tables

Every drag-and-drop email builder, including the one inside GoHighLevel, outputs nested table elements. That is simply how you get columns and aligned blocks to render consistently across mail clients. It is also the loudest structural tell that a message came out of a marketing platform, because nobody writing a genuine note to a colleague produces a table layout. Our email had them throughout, wrapping every section.

Signal 2: There Was a Tracking Pixel Inside It

Open tracking works by embedding a one-pixel transparent image and watching for the request when the message renders. Most platforms add it by default, ours included, and most senders never consciously choose it. The pixel is invisible to your reader and completely visible to the receiving mail server. A message carrying one is, by definition, a message somebody is measuring at scale.

Signal 3: Every Link Had Been Rewritten

Click tracking replaces your real URL with a redirect through the sending platform's tracking domain. Your reader sees your words. The receiving server sees a link pointing at a marketing vendor rather than at your clinic's domain. Layer UTM campaign parameters on top and you have attached campaign metadata to a message you wanted to read as personal. Ours had both, on every link.

Signal 4: It Looked Like a Brochure

Styled call-to-action buttons, meaning an anchor tag wearing a background color, padding and rounded corners. Background colors on containers. A block of CSS in the head. Decorative type sizes and colors. Any one of these is a reasonable design decision on a landing page and works against you inside an inbox. A high image-to-text ratio compounds it, and any image wider than a couple hundred pixels reads as a banner rather than a screenshot.

Signal 5: The Sender Read Like a System

It came from a support address. Support addresses are where ticket confirmations and password resets come from. They are not where one practitioner writes to another. The from-name was the brand rather than a person. Between those two fields alone, the receiving mail client had no reason to treat the message as personal correspondence, and neither did the human who glanced at it.

Signal 6: All 244 Went Out at Once

A single campaign blast delivers an entire list at one timestamp. That is a bulk-sending pattern and it is trivially detectable. Real one-to-one email does not arrive in a synchronized wave, and a sending pattern is one of the few signals no amount of HTML cleanup will disguise.

What We Sent Instead

The rebuild was not a redesign. It was a demolition. The follow-up emails were written as raw HTML inside a workflow email action rather than assembled in the visual builder, and the entire permitted toolkit was:

• Paragraph tags and line breaks
• Bold and italic for emphasis
• One inline font-family and font-size on a single wrapper element
• A personalization token for the recipient's first name
• Plain text URLs, typed out rather than hyperlinked, so nothing could be rewritten

Removed: every table, every image, the tracking pixel, the click-tracking wrapper, the UTM parameters, the styled button, the container background colors, and the CSS block.

Changed: the sender moved from a support address to a named personal address, the from-name became the doctor's name, and delivery moved from one campaign blast to a workflow that enrolls 40 to 50 contacts per hour, spreading sends across roughly five hours instead of one event.

Added: an explicit invitation to reply. "Reply to this email and I will set it up for you" does two jobs at once. It gives an interested reader the lowest-friction way to respond, and a reply is the strongest engagement signal a mailbox provider can observe about a sender.

One thing we did not remove, and neither should you: the compliance elements. Stripping design is not the same as stripping legal obligations. The FTC's CAN-SPAM compliance guide requires commercial email to carry a clear opt-out mechanism that remains functional for at least 30 days, to honor opt-out requests within 10 business days, to include a valid physical postal address, and to avoid false header information or deceptive subject lines. Google's bulk sender rules require the unsubscribe headers on top of that. Those stay in. Cut the decoration, never the disclosure.

The Trade Nobody Mentions

Turn off open and click tracking and you lose your open rate and your click rate. That is a real cost, and you should choose it deliberately rather than discover it three weeks later when someone asks for a report.

What we track instead:

• Replies, which on a list this size is the only number that ever mattered. A 244-person provider list does not need a 22% open rate. It needs eleven conversations.
• Workflow enrollment inside the CRM, which shows exactly who received the message and when, with no pixel required.
• A tag applied automatically on reply, so the pipeline advances without anyone watching an inbox.
• Analytics on the destination page, since your website can carry its own tracking even when the email carries none.

If you are mailing a 40,000-person consumer list a seasonal offer, keep your tracking and accept Promotions. That is where an offer belongs, and Google built annotation tooling precisely to make deals display well there. If you are a clinic writing to a few hundred providers or patients you actually know, the placement is worth more than the metrics.

What This Does Not Fix

Container problems and content problems are different problems, and it is expensive to confuse them.

A perfectly plain email to the wrong list still produces nothing. So does a plain email with no relationship behind it, or one that asks for a several-thousand-dollar decision in the opening paragraph. Our post-mortem landed on formatting because the list was warm and the offer was sound, which meant formatting was the remaining variable. Fix formatting first only when you are confident those two are already true. Otherwise you will strip out every table, hear nothing back again, and conclude wrongly that email does not work for your clinic.

The same discipline applies to the campaigns that fill your schedule. Patient email built to fill a launch event needs dates and a reason to act now, not a better template. More on filling a launch event calendar. And the rules governing what you may claim do not relax because the email is plain text, which is worth reading in full before you write a single subject line. More on marketing without overpromising.

How to Test This in Twenty Minutes

Before the real list ever sees a message, run this:

• Send the rebuilt email to a personal mailbox that is not on your clinic's domain, and to a second free mailbox if you have one.
• Note which tab it lands in.
• Open the message, view the original source, and confirm your authentication passed and that no tracking pixel or rewritten link survived the build.
• Reply to yourself once from that test account. Engagement history between two addresses is real and it accumulates.
• Only then enroll the live list, in hourly batches rather than one send.

Repeat the test every time you change sending platforms, sender addresses, or templates. Inbox placement is not a setting you fix once. It moves with your sending behavior, and outcomes vary by domain and recipient.

The Practical Takeaway

Most clinics diagnose a silent email list as a copy problem and rewrite the subject line. Rewrite it a dozen times and it will not matter if the message is being filed where nobody looks. Before you touch a word of the copy, open the last email you sent and ask a simpler question: does this look like a person wrote it, or does it look like software generated it? The answer is usually visible in about four seconds, and it explains more about your reply rate than the headline ever will. The device is not the problem, and neither is the sentence. The system around it is. More on the systems that compound patient value.