Compliance · 6 min read
A Mental Health Coach Is Not a Therapist, and DMs Blur It
On this page
- The problem is that your inbox does not know the difference
- The rule that comes before every other rule
- A blocklist is necessary and not sufficient
- Never let a template describe you as something you are not
- The crisis route must not depend on the automation
- The instant acknowledgement is a claim about attention
- What is honestly automatable, and the ending
This post is for a non-clinical coach: somebody working on habits, stress at work, confidence, sleep routines, burnout. Not a licensed clinician.
If you are a therapist or a counsellor in practice, stop here and read the therapists post instead. It owns that case, it is more thorough than anything on this page, and the posture it argues for — automate almost nothing, get out of the way fast — is the right one for a clinical inbox.
The problem is that your inbox does not know the difference
You are careful. Your captions say coach, not therapist. Your bio does not claim a qualification you do not hold.
None of that survives contact with the inbox. People message a mental health coach the way they would message a therapist, because from the outside the distinction is invisible and, at the moment somebody writes, irrelevant to them. They describe symptoms. They ask whether to stop a medication. They ask if what they are experiencing is normal. They write at 2am.
So the boundary you maintain carefully in your content has to be maintained mechanically in your replies, by a system that has no judgement at all.
A template does not know it is not a clinician. It will answer with the same confidence either way, which is exactly the risk.
The rule that comes before every other rule
Never automate a reply to anyone expressing distress or crisis. Not a warm one, not a signposting one, not one you wrote carefully on a good day.
When such a message arrives, the automation's only job is to send nothing and put the conversation in front of a human being as fast as possible. That is not a trade-off to balance against response times or conversion. It is the requirement the rest of the build is arranged around, and an automated reply there does not read as efficiency — because it arrives instantly and sounds attentive, it reads as a person having received the hardest message somebody has written all year, when nobody did.
A blocklist is necessary and not sufficient
Write the negative keyword list. Then understand what it will not catch, because relying on it as a safety system is the mistake this section exists to prevent.
Distress does not arrive in the vocabulary a filter expects. It arrives as four lowercase words with no punctuation. As "hi" at 2:40am followed by nothing, then three paragraphs at 3:15. In Hinglish, in metaphor, as a story reply about something unrelated. Matching folds case and punctuation and matches whole words — it does not read meaning.
So invert the posture. Instead of replying to everything except a list of exceptions, build an automation that replies to nothing except a short list of situations you decided in advance were safe.
- Scope every automation to specific posts. Under Advanced, name them. One workshop announcement, one post where you asked people to comment a word. Never the whole account.
- Make the trigger a word you asked for. Something typed on purpose because your caption told them to — not a broad any-of list that will match a sentence somebody wrote in pain.
- Keep the reply to a resource and a route. The worksheet, the page, how to enquire. Nothing responding to what they said, nothing reassuring, nothing interpreting.
- Layer the blocklist on top anyway. Second net, never the first.
- Default everything else to silence. A message that matches nothing gets no automated reply and a person reads it. That is the correct outcome, not a coverage gap.
Never let a template describe you as something you are not
An automated reply written to sell will reach for clinical weight, because clinical weight is persuasive. "Heal your anxiety." "A proven protocol for depression." Those sentences position a coach as a practitioner, and they are the ones a generated reply produces most readily, because that is how the internet writes about this subject.
The template has to say plainly what you do and what you are not. One sentence, in your own words, that a reader could not misconstrue. It will filter your inbox better than any keyword list, because the people who needed a clinician now know to look for one.
The crisis route must not depend on the automation
Your account needs a visible, current route for somebody who needs help now, and it has to work when the automation is off, misconfigured or blocked by a check. So it lives where nothing has to fire: the bio, a pinned post, the website.
Publish your own. Services differ by country and they change, and a stale entry is worse than none, so this post will not print any. Pick the ones relevant to where your audience is, and check them twice a year. Say in the same place that Instagram messages are not monitored continuously and are not a route for urgent help — where somebody reads it before they write, not in a reply afterwards.
The instant acknowledgement is a claim about attention
The most tempting automation here is the kindest-sounding one: "Thanks for reaching out, I'll get back to you soon." At 3am, to somebody who has just written something difficult, that sentence claims a person received it. Nobody did, and because it landed instantly it reads as the most attentive message in the thread.
If one exists at all it has to be honest about being automatic and promise nothing about when a human will read. Most people, having written that out, would rather send nothing.
What is honestly automatable, and the ending
Safe, because it is administrative and identical for everybody: how to enquire, whether sessions are online or in person, which languages you work in, how long a session runs, where to find something you have published. That is close to the entire list.
Not safe, and not close: anything responding to what somebody described, anything reassuring, anything interpreting, anything about medication, and anything about whether a person should seek help urgently.
Takeover stands the automation down in a thread the moment you reply by hand, permanently. Use it constantly. Templated replies are free and unlimited; a credit is spent only when the AI writes a new reply, which on this build should be close to never. Free tier is 100 credits with no card, packs from ₹499, and cost is not the deciding factor here.
No calendar, no booking, no payment link, no CRM sync, no WhatsApp, no sequences. Leads leave as a CSV. Official Graph API only: seven days per comment, twenty-four hours inside a DM thread, restarted only by their message — a platform constraint, not a model of care.
Running no automation at all is a legitimate configuration. If nobody has time to read Activity weekly and re-read the templates quarterly, off is correct. Wellness brands face a nearby version of this from the commercial side.
