Key takeaway
A dental practice runs two email programs rather than one, and the line between them is drawn before the software sees it: a message about a patient's own care isn't selling, and one promoting somebody else's product, or paid for by a brand, is. Both halves leave the same domain on one reputation. I'd worry less about the whitening promotion than about it costing you the recall reminder.
What a dental practice's email program is actually for.
Email marketing at a dental practice is scheduled mail sent to people who gave you their address, from a domain the practice controls. The six-month recall, the note after a crown, the whitening offer in November.
What that description hides is that a practice isn't running one email program. It's running two, and the two halves have almost nothing in common. One half is about a patient's own mouth: the recall that's due, the post-op check, the treatment plan that was quoted and never booked. The other half is selling something: whitening before wedding season, aligners, the ask for a referral. Both leave the same account on the same afternoon, and the advice a practice picks up runs the two through one set of rules, which is why it fits neither.
The failure I've watched most often isn't a legal one. A practice sends the elective-treatment blast to the whole list, takes the complaints it earns, and three weeks later the recall reminders are landing in spam. Nobody catches it, because a recall reminder that doesn't arrive looks exactly like a patient who didn't book.
Two programs, and the line is drawn before your software sees it.
It's tempting to treat this as a purchasing question: find the platform that handles patient data properly, and the problem is solved by an invoice. It isn't a property a product has. A dashboard-first product can only apply a rule at the moment somebody opens the screen, and this one has to hold on every send, including the ones nobody is sitting in front of. Two questions decide which program a message belongs to: was a patient's own record used to choose who receives it, and is the message selling something.
The first question is easier than practices expect, and the answer is uncomfortable. Clinical detail isn't the threshold. The fact that a named person is a patient of a named dental practice is already the confidential part, so a marketing list at a practice is a list of exactly that, whatever the fields on it are called.
Which means two sends can leave the same account on the same day, drawn from the same list, and only one of them needs a patient's say-so first. The sorting happens before the software does, and it's done against a line that's short enough to hold in your head.
The other thing both halves share has nothing to do with any of that. It's the domain they leave from, and it's what decides whether either of them is seen.
What counts as selling.
The line is one sentence long: a message is selling when it encourages the reader to buy or use a product or service. Three cases fall out of it, and each turns on what the message is about rather than on how carefully it's worded.
- About that person's own care: their treatment, care coordination, an alternative treatment worth considering, or a health service the practice itself provides. Not selling, as long as nobody outside the practice is paying for the send.
- About a product or service the practice doesn't itself provide: that's selling, the list that chose who received it is patient data, and that pairing is where you need the patient's say-so first.
- About somebody else's product, and they paid you: payment from the brand whose product is described. It changes the answer whatever else the message is.
That line runs counter to what a practice usually arrives believing. The recall reminder handled as the delicate case is a message about the patient's own treatment, and it sits outside selling entirely. So does the whitening promotion, as long as the practice performs the treatment itself and nobody outside it is paying for the send. The case that catches practices out is the third: an aligner or whitening brand that pays for the send, or supplies the offer and the co-branded creative for a share of it. The money reaches a message that would otherwise fall on the safe side.
Where the patient's say-so actually bites.
The requirement is short, and it has almost no exceptions: a practice needs the patient's explicit permission before using their record to send them marketing. The two carve-outs are a conversation face to face and a promotional gift of nominal value. Neither of those is an email, and the phrase "HIPAA-compliant email" a vendor puts on a pricing page is usually standing in for that sentence without anybody having read it.
One thing rides on top of it. Where a third party is paying for the send, the permission the patient gives has to say so. That disclosure is a term of the permission itself, not a line somewhere in the footer of the message.
Two staples of the dental newsletter sit inside this. A patient's account of their own treatment and a before-and-after photograph of their mouth, used to promote the practice, are a use of that patient's record to sell, and the requirement reaches them exactly as it reaches a discount code.
The list in somebody else's hands is a separate question from the message in it. Anyone who holds, moves or stores that list on your behalf is holding patient data, and whether a given arrangement counts turns on what the file identifies rather than on the feature list around it. It isn't answered once, either. Every send gets sorted again against the same two questions, and a sorting rule that applies on every send is something a system can hold rather than something a person re-decides at the point of sending.
The gate that decides whether any of it arrives.
That gate is published by the mailbox providers about their own systems. Nobody outside them sets or enforces it, and it's the only one of the two that touches every message a practice sends. I'd put the 0.3% complaint ceiling on the practice manager's monthly list before anything else on this page, because it's the number that decides whether the recall reminders land. Google's sender guidelines and Yahoo's sender best practices agree on nearly all of it, and which part applies turns on how much you send.
- Every sender, whatever the volume: SPF or DKIM on the sending domain, valid forward and reverse DNS, and a TLS connection. One-time work, and the difference between mail that authenticates and mail that looks forged.
- Any marketing message, whatever the volume: a spam complaint rate held below 0.3% in the provider's own reporting. Both publish that ceiling for every sender, and it's the number a send cadence is really tested against.
- More than 5,000 messages a day to Gmail accounts: SPF and DKIM and DMARC together with the From domain aligned, plus one-click unsubscribe supported in the message headers. Yahoo asks the same of the senders it calls bulk.
Two things in that list are worth a second read. Google counts that threshold per day, across everything the domain sends to Gmail accounts, rather than across a month or across a whole list, so a practice picturing a bulk sender as somebody else entirely can find one newsletter to a list in the thousands, most of it on a single provider, sitting closer to the line than expected. And the two providers don't agree on the unsubscribe deadline: Yahoo asks the senders it calls bulk to action one within two days, which is the tighter number and therefore the one to build to. When I ran a business on this, our own automatic flag sat at 0.5% complaint rate, which sounds cautious until you notice it's already past the 0.3% ceiling above: by the time an account tripped it, mail had likely started landing in spam already.
One domain, two jobs, one reputation.
Reputation attaches to the sending domain. Not to the campaign, not to the individual message, and not to the half of the program the practice considers important. The whitening promotion and the recall reminder leave the same domain, so they're scored together.
That's where this vertical's stakes sit. The elective-treatment blast is the send most likely to earn complaints, because it goes to everybody and asks for the most. The recall reminder is the send that carries most of the program's value, and it's the one that quietly stops arriving when the score drops. A practice that never sends a promotion never meets the problem. A practice that sends one a month and doesn't watch the complaint rate is trading recall bookings for whitening inquiries without ever deciding to.
The remedy is a standing condition rather than a purchase. The two jobs leave from separate subdomains, so the promotional score and the transactional score aren't the same score. Both stay authenticated, and no send goes from the wrong one. The complaint rate on the one that goes widest is checked every time it goes, and a rise in it is read as a delivery forecast for the other one.
I've almost never seen this reported as a deliverability problem. It gets reported as the recall list going cold, or as a hygiene schedule with gaps in it nobody can account for, and by then the template has usually been redesigned twice. It stays invisible because keeping the split intact is small recurring work with a fixed answer every time: two domains authenticated, two reputations watched, every send leaving from the right one. That work gets skipped when it lives behind a dashboard somebody opens once a quarter.
What a practice sends, sorted by which half it's in.
Grouped this way rather than by which send performs best, the inventory is shorter than the usual list of ideas, and the grouping is most of the work.
- Recall and reactivation: about the patient's own care, outside the definition of marketing, and the send that carries most of the program's value.
- Post-visit and treatment-plan follow-up: the same half, and the one where a subject line naming a procedure does the most damage in a shared inbox.
- The practice newsletter: mixed by nature, which is what makes it the hardest to place. Whichever half it lands in is decided by what's in it, item by item.
- Elective and seasonal promotion: the other half entirely, the send that needs the most thought before it goes, and the one putting the domain's reputation at risk.
Keeping the two apart, month after month.
Two kinds of send, two reputations, one domain, and a standing job underneath them: authenticate both, keep every send on the right one, and watch the complaint rate on the one that goes widest.
That's the work Nitrosend is built for: MCP-first, so every capability is an API endpoint and an MCP tool before it's a screen, and those checks are something an agent runs rather than something a person remembers. Nitrosend specializes in email and doesn't enter business associate agreements, so a send whose list identifies who a practice's patients are belongs with a vendor that does. The free tier is enough to stand that split up: two subdomains authenticated, and the wide one watched.
Sources
- Google, email sender guidelines: SPF or DKIM for every sender, forward and reverse DNS, TLS, the 0.3% spam-rate ceiling, the 5,000 messages a day bulk threshold, and one-click unsubscribe.
- Yahoo, sender best practices: SPF or DKIM at a minimum, both plus a DMARC policy for bulk senders, a spam rate below 0.3%, and unsubscribes honored within two days by bulk senders.
Common questions
Compliance isn't a property an email platform has, so no product answers this by itself. What decides it is the send: was a patient's own record used to choose who receives this message, and is the message selling something. A recall reminder is about that patient's care and clears on the first question. A promotion for a brand that paid you to send it fails on the second, and needs the patient's explicit permission first.
Yes, with a distinction that matters. A promotion for treatment the practice performs itself, sent to your own patients with nobody outside paying for it, sits on the safe side of the line. A promotion for somebody else's product, or one a brand paid for, is the case that needs each patient's explicit say-so before their record is used to target it. Both, either way, need a working one-click unsubscribe.
No. A recall reminder is a message about that person's own care, which is exactly what sits outside selling. That's the opposite of what most practices assume, and it's why the delicate handling usually lands on the wrong message: the recall is fine, and the third-party promotion in the same newsletter is the one to stop and think about.
The two surfaces that leak most sit outside the message body. A subject line renders in a preview pane on a screen a spouse, an employer, or a family tablet may be looking at, and one naming a procedure says what the appointment was for before anybody opens anything. A segment name is the other: a segment called implant consults is a record of who had one. Clinical detail was never the threshold, because the fact that somebody is a patient of a named dental practice is already the confidential part.
Anyone who holds, moves or stores your patient list on your behalf is holding patient data, whatever the product is called, so the agreement is the normal expectation rather than an unusual ask. Whether a given arrangement counts turns on what the file actually identifies, not on the feature list around it. A vendor that can't have the conversation is answering the question.
A reminder for a booked visit is a message the patient's own action asked for, so it isn't a promotion and doesn't have to read like one. Google and Yahoo draw no such line at the domain level, though: they ask for one-click unsubscribe support on marketing and subscribed mail, and they judge the domain on its complaint rate whatever you call the message. So the reminder needs no unsubscribe, and the newsletter it travels beside decides whether the reminder arrives.
There's no cadence number worth quoting, because the constraint isn't frequency, it's complaints. Google and Yahoo both publish a spam-complaint ceiling of 0.3%, and that's the figure a sending schedule is actually tested against. A monthly promotion to a list that hasn't heard from the practice in a year will cross it faster than a weekly one to people who booked last month. The practical version: watch the complaint rate on whichever send goes widest, and let it set the schedule rather than the other way round.
Two questions come before any feature list, and they're about the practice's own sends rather than the software. Does the list being uploaded identify who the practice's patients are, and is a given send about that person's own care or about selling something. Those two answers decide what the platform is holding and what kind of arrangement that implies. Then there's the part that's true whoever sends: the mail leaves a domain, reputation attaches to that domain rather than to the campaign, and a promotional send and a recall reminder from the same domain are scored together.