Patient messaging & SMS
Fertiligent gives your clinic a single place to hold two-way conversations with patients (in the portal, over text (SMS), and by email, which your team can now start as well as reply to) with optional AI assistance that drafts replies for your team to review. This page covers how messaging works and how the AI fits into your day.
Conversations
Messages are organized into threads, so a conversation with a patient stays together over time. Within a thread your team can:
- send and receive messages, with file attachments;
- reuse templates for common replies, so answers stay consistent;
- see read status and who said what;
- send a bulk message to many patients at once (for example a clinic-wide notice).
Patients can reach your staff and admins; staff can message anyone. Everything lives in one inbox in the portal.
AI-assisted replies
When AI messaging is enabled, incoming patient messages can be analyzed and a suggested reply drafted for your team. The AI is built around a review queue and a simple risk assessment so a person always stays in control:
- Risk level. Each message is assessed as safe, needs review, or escalate. Anything sensitive is routed to a human rather than answered automatically.
- Review queue. Suggested replies land in a queue where staff can approve, edit and send, reject, re-run the analysis, or escalate to a colleague.
- Draft assist. Even outside the queue, staff can ask for a draft reply to a thread and edit it before sending.
- Pause per conversation. You can pause AI on any single thread, for example a delicate conversation you want to handle entirely by hand, and resume it later. Claiming a conversation in Conversations sets the same pause, and it holds across every channel including text: once a person on your team owns the conversation, the assistant does not answer that patient until somebody resumes it. Carrier keywords are the exception, STOP and START always take effect, whoever is holding the thread.
AI modes
How much the AI does is controlled per clinic by a mode:
| Mode | What happens |
|---|---|
| Off | The AI does no drafting; staff handle every message. |
| Review queue only | The AI drafts replies, but nothing is sent until a staff member approves it. |
| Safe auto-send | Low-risk, high-confidence replies can be sent automatically; anything else still goes to the queue or a human. |
Whichever mode you use, sensitive messages are escalated to a person, and every AI action is recorded so you can audit what happened.
Text messaging (SMS)
Patients can also reach your clinic by text, and the same assistant that answers your phone line handles the conversation. A few things are specific to SMS:
- Identity and consent come first. Before any personal health information is shared over text, the patient must both verify their identity (by date of birth) and give explicit consent to receive health information by text. A verified patient who hasn't consented gets a short message asking them to opt in, never their private details.
- Opting in and out. Patients can reply STOP to opt out at any time (which also withdraws their consent to receive health information by text) and START to opt back in. Replying HELP returns help text. These are handled automatically.
- Nothing gets lost. A text that doesn't answer an open question, and that the assistant isn't set up to hold a conversation about, still gets a courteous auto-reply pointing the patient to call or use the portal, rather than vanishing.
Asking a patient how they'd like to be reached
An outreach sequence can send a preference question: a confirmation that the clinic has received the patient's referral, followed by a short menu.
Reply 1 for text messages, 2 for email, 3 for a phone call, or 4 if you no longer wish to proceed. Reply STOP to opt out.
The patient can reply with the digit or in their own words, "text", "email" and "call" are all understood. A reply the assistant can't interpret gets a re-prompt offering the same options, never a different set. The stated preference is recorded against the prospect along with where it came from, so your team can see the patient chose it rather than a coordinator assuming it.
Your clinic authors the wording of this message. The meanings of the four options are fixed by the platform, so a reworded menu can't quietly change what "4" does.
How a patient withdraws
Withdrawal is deliberately harder to trigger by accident than an opt-out, because withdrawing means somebody has to tell the referring physician.
| What the patient sends | What happens |
|---|---|
| 4, while the preference question is open and it is the last step in the sequence | Recorded as a withdrawal. The prospect moves to Withdrawn, outreach closes, and any stated contact preference is cleared. |
| 4, while the question is open but further steps remain | Routed to a coordinator to handle, rather than closing the file automatically. |
| A withdrawal in their own words: "not interested", "stop contacting me" | Nothing is recorded. Outreach is put on hold so the patient stops receiving messages, and a coordinator is alerted to confirm. |
| 4, with no question open | Nothing is recorded. A lone digit only means "withdraw" in the context of the menu it answers. |
STOP is a carrier-level opt-out: it stops the texts. Option 4 is a care decision: the patient is telling the clinic they no longer wish to proceed with their referral. One does not imply the other, and they are recorded differently.
An answer to your own question is not a withdrawal
If a coordinator texts a patient a question ("reply YES if you'd like to go ahead, NO if not") and the patient replies with a bare "no", that is read as the answer to the question you asked. It is no longer treated as the patient withdrawing their consent to be contacted.
The rule has two halves, and both have to hold: the reply carries no words about consent of its own, and your clinic spoke last on that number within the past few hours. A message that speaks for itself ("please stop texting me", "I withdraw my consent") is still recognised as a withdrawal whether or not you spoke last, and STOP is handled before any of this and always works.
Before this, a bare "no" answering the clinic's own question could be escalated as a withdrawal of consent, because nothing in the system knew the clinic had asked anything.
A text that asks for nothing doesn't get a consent request
The consent request goes only to a patient who asked for something. A text that carries only an email address or phone number, usually the answer to something your team asked, is not read as asking for health information by text. It goes to a coordinator rather than prompting the consent question. The same holds for a message that asks for nothing at all: a one-word typo correction, thanks, ok.
Your team can also send a text to a patient or prospect, for example a referral-received confirmation, from the portal, subject to the same consent rules. The confirmation a prospect gets when an appointment is booked follows them too: it goes by text only to a number cleared for clinical texting (SMS consent counts whichever linked record it was saved against), it starts with your clinic's brand name, and it always fits in a single text. See The booking confirmation. Text conversations are logged alongside calls, portal messages and email in the Conversations worklist, so your team reviews everything in one place: see Conversations.
Inbound replies are also written to the prospect's own timeline, so a coordinator looking at the record sees what the patient actually said rather than only that a reply arrived.
Every outgoing text carries what the carrier reported underneath it, delivered, not delivered, or delivery not confirmed where no report has come back either way. That last one is not the same as "not sent"; see Conversations.
What the assistant can do over text
Over SMS the assistant runs the same AI as your phone line, with a tool set tuned for texting. Depending on what the patient needs (and, for anything personal, once their identity is verified and consent is on file) it can:
Answer and send information
- Answer questions about your clinic from your knowledge base: hours, location and directions, parking, services, pricing, and policies.
- Text a patient the public information they ask for, such as an address, how to prepare for a visit, or a public link. (Public info only, never private records or results.)
Book an intake call
- Find open appointment times and book a nurse intake call on the spot. See Intake nurse availability.
Help a verified patient with their own care
- Look up their upcoming appointments and the status of recent results.
- Check a referral's status, share the referred physician's current wait, and offer a shorter-wait alternative.
- Confirm an upcoming appointment, or request to cancel or reschedule one.
- Request to switch to the alternative physician it offered.
- Update a contact detail, or provide a missing referral detail (health card, email, or mobile number).
Always
- Capture a new inquiry as a prospect, with the patient's consent.
- Hand off to your team the moment something needs a person.
For anything that touches a patient's record or the calendar (a reschedule, a physician switch, a contact-detail change, a missing referral field) the assistant records a request or draft for your staff to review and apply. It never edits the record or moves the calendar on its own. The one exception is booking a nurse intake call, which is confirmed on the spot.
When the assistant should say nothing
Separately from the modes above, your clinic can require the assistant to use only approved wording over text and phone. This is the approved replies setting, and it has two positions:
| Setting | What happens to a message the approved wording doesn't cover |
|---|---|
| Free (default) | The assistant answers in its own words. |
| Strict | Over text, portal and email, nothing is sent: the message is recorded on the conversation and raised for a coordinator to answer. That holds however the message reaches a coordinator, including when the assistant hands it over on its own. On the phone the caller still hears a line, because silence on a call is not a hand-off. |
At Strict the assistant used to reply with a holding sentence ("Thanks, we've passed this to our team"); which could not relate to what the patient had actually said. A patient answering one of your questions with "YES" received it. That reply is gone; the hand-off to a coordinator is not. A portal thread held back this way shows Escalated in your team's message list, so it reads as waiting on a person.
On a Strict call, and with Eva on your website, the hand-off line is said once. A second question the approved wording doesn't cover gets a shorter line saying it has been added to the message for the care team. Both lines can be reworded on the Approved Replies settings tab. On a call, the pause while the assistant works out an answer is filled only with the fixed line "One moment." See Approved wording on calls.
Emergency wording, consent, identity checks and the carrier keywords are approved wording, so they still answer at Strict. A patient who texts STOP is opted out and told so, exactly as before.
A text from a number your clinic doesn't recognise is likewise recorded for staff without an automatic reply going back.
A reply the assistant withheld is shown as withheld, with the reason, in the Conversations worklist and the conversation history; it never looks like a reply that was sent. A factual readback of a released result is not treated as medical advice and is not held.
The email a patient receives
When a message is waiting for a patient in the portal, the notification they get by email is deliberately plain and deliberately empty of clinical content:
- It greets them by first name and its subject reads Jane, you have a new message. It says a new message is waiting in their patient portal, shows the sign-in address beside the button, and says who sends it and on whose behalf. The message itself is never in the email.
- The words "secure message" do not appear; they are the phrase spam filters object to, and a filtered notification helps nobody.
- A patient who has not yet read the last such email is not sent another when a further message arrives. The next one goes once they have read it, or after a day.
- Where your clinic has it switched on, notifications leave from a dedicated Patient Messaging address on a sending domain reserved for patient mail, kept apart from staff mail, and their delivery reports (delivered, bounced, filtered as spam); are recorded, so a message that never arrived can be traced.
Availability
Messaging, AI-assisted replies, and SMS are each enabled per clinic. Because text messaging with real patients depends on identity verification, consent, and the required carrier agreements being in place, SMS is rolled out in a controlled way. Ask your Fertiligent contact about enabling it for your clinic.
Related
- Voice phone assistant: the same assistant on the phone.
- Eva, the patient companion: chat on your website.
- Knowledge base: what the assistant draws its answers from.