What parents worry about

Self-harm alerts: what they catch, and what no app can do

If something is happening tonight, the free helplines are at the top of this page. If you are here calmly, to understand what a self-harm alert really is before you need one, this guide explains what such a check can catch, what it can never see, why it exists only on Android, and what to do when one arrives.

If it is happening tonight

If you are reading this because of something you have just seen, stop reading and go to your child. Stay calm, stay with them, and use these numbers. They are free, open now, and the people who answer do this every night.

Outside the UK, call your local emergency number or your country's child helpline.

Then read what to do when your child sends a message about hurting themselves, which walks through the next hour. The page you are on now is the calm version, for a parent who wants to understand these alerts before ever needing one.

The short answer

A self-harm alert is a check that runs on a child's own phone. It looks at messages as they arrive for language suggesting someone is being hurt or wants to hurt themselves, and when it finds something it tells the parent: the kind of danger, how serious it looks, the app it came from, and a few words of it.

It is not a transcript, and it is not a diagnosis. It is a nudge to go and talk to your child tonight instead of next month — nothing more than that.

What the check actually looks at

It reads the text of an arriving message, the words that land in a notification, and compares them against patterns of worrying language. Broadly, four kinds:

That last one matters more than parents expect: a great deal of what a check catches is not your child in danger, but your child holding somebody else's crisis, alone, at eleven at night.

A check built properly runs on the device, so the message never travels anywhere; keeps only a short extract, a line at most; and shows on the child's own screen that it is on, so nothing runs in secret.

Why it only exists on an Android phone

Apple does not let any app read another app's messages or notifications on an iPhone. That is deliberate, and there is no setting or clever way round it. So on an iPhone there is no message-safety alert of any kind, from anybody — if an app claims to watch iMessage or WhatsApp there, treat the claim as untrue.

Android does allow it, through a permission the child grants and can take away again. That is why this is an Android story, and why it still reaches most of the world's children: about 70% of the world's phones run Android (World Population Review, using StatCounter data), rising to 96% in India and 94% in Pakistan.

What it cannot see

That is not a short list, and no honest maker hides it. An alert is a smoke alarm: worth having, but not a fire brigade, and it does not watch every room.

False alarms, and why they are the price

You will get alerts that turn out to be nothing. Song lyrics. A set text for English. A friend's dark joke. A game where somebody dies. A system tuned never to cry wolf would also miss the one that mattered, so the sensible setting is the one that occasionally embarrasses you.

Treat every alert as an invitation to check in, never as evidence. "I got a message that made me want to see you were all right" is a fine thing to say when nothing is wrong. If your child rolls their eyes at you, that is a good evening.

Consent, and why hiding it backfires

Your child should know the check is on. They should read a screen that says what is looked at, what you will see, and what is never kept — and "not now" should be a real answer that changes nothing. That sounds like a weakness. It is the opposite.

The point of an alert is to start a conversation, and a conversation is hard to start from a position your child experiences as betrayal. Secret monitoring, once discovered, usually ends the confiding altogether.

There is a plainer reason too. Two in five children who saw the worst thing they had ever seen online told nobody at all (LGfL research, via Internet Matters). Of children who had been groomed and told someone, 93% told a friend and only 19% told their mother (Bravehearts). Being the parent your child might tell is worth more than any alert. The alert is for the night they do not.

When an alert arrives

  1. Sit down first. Two minutes. Panic is what makes the next hour worse.
  2. Do not lead with the phone. "I got an alert" turns it into surveillance. Go and be with them, and start with an ordinary question.
  3. If it points at real danger, ask directly. "Are you thinking about ending your life?" Then stop and wait. Asking does not plant the idea; mental-health services advise asking.
  4. Stay tonight. Sit with them, sleep in the room, do not leave them alone.
  5. Ring the GP in the morning and use the words "my child has talked about suicide" or "has been self-harming". Those words get an urgent appointment.
  6. If the message came from a friend, that child needs an adult too. Help your child tell one.

The hour-by-hour version is in our crisis guide. Bullying and self-harm often arrive together; that picture is in our guide to online bullying.

The numbers

WhatThe numberSource
UK 17-year-olds who have self-harmedAbout 1 in 4; 7% with suicidal intentUCL Millennium Cohort Study
UK 11–16-year-olds with a probable mental disorder13% in 2017, about 23% in 2023NHS survey, via YoungMinds
Childline sessions about suicidal thoughts, 2024/2518,981 in the year — about 52 a dayNSPCC
Worldwide, aged 10–19, with a mental disorder1 in 7; suicide is the 4th leading cause of death at 15–19WHO

One more, from ACAMH: young people who self-harm usually tell friends rather than parents, and between a third and a half get no help until it becomes an emergency. That gap is the whole reason anyone builds an alert.

What to say to your child

This is the conversation for a calm evening, not a crisis. "I have turned something on that checks messages arriving on your phone for the kind of thing that means somebody is being hurt. I cannot read your chats and I cannot see who you talk to. If it ever tells me something, I will come and ask you, and I will not be angry. I would rather know and be wrong."

About this guide

This guide is published by the parent-led team behind Safeinest, a family-safety app. It is written to stand on its own: nothing above needs the app. If, later, you want to know what an app can and cannot do on an Android phone or an iPhone, the honest version is on our features page.

Common questions

What is a self-harm alert?

It is a check that runs on a child's own phone, reading messages as they arrive for language about hurting oneself, hopelessness or wanting to die. When it matches, the parent gets a short alert: the kind of danger, how serious it looks, the app it came from and a few words. Not the conversation, and not a diagnosis.

Do self-harm alerts work on an iPhone?

No, and nobody can offer them there. Apple does not allow any app to read another app's messages or notifications on an iPhone, so no message-based alert is possible, whatever the advertising says. If a product claims to watch iMessage or WhatsApp on an iPhone, treat that claim as untrue.

Can an app tell me if my child is suicidal?

No. An app can only notice words that arrive on a phone. It cannot read a mind, and a child who is hurting quietly and types nothing is invisible to every app ever made. What an alert can do is shorten the time between something being said and a parent knowing — worth having, but not a safety net.

Will my child know the check is switched on?

They should, and on a properly built app they do. Your child reads a screen explaining what is looked at, what you will see and what is never kept, and declining is a real answer that changes nothing. Hidden monitoring, once found, usually ends the confiding that keeps children safe.

Does it read all my child's messages?

No. The check runs on the phone itself, looks at text as it arrives, then forgets it. What reaches the parent is a category, a severity, the app name and a short extract of at most a line. There is no feed of conversations and no list of who your child talks to.

What if it is a false alarm?

Expect some. Song lyrics, a set text for English, a friend's dark joke and video games all set alerts off. A system tuned never to cry wolf would also miss the real one, so a few unnecessary conversations are the price. Treat an alert as a reason to check in, not as evidence.

Does it work in Urdu, Hindi, Arabic or Bangla?

Not yet. The words the check understands are English only today, even though the apps themselves are translated into many languages. In a household that messages in another language, an alert will miss things, and it is better to know that than to feel covered. Local-language detection is being worked on.

Can it see WhatsApp voice notes or photos?

No. Spoken words are never checked, so a voice note or a call passes unseen, and a picture is only a picture to the check, including a screenshot of a message. It also cannot see what your child types out and sends themselves, or a chat in an app whose notifications are muted or switched off.

What should I do the moment an alert arrives?

Sit down for two minutes before you move. Then go to your child in person and start with an ordinary question rather than the alert. If it points at real danger, ask plainly whether they are thinking about ending their life, stay with them tonight, and ring the GP in the morning saying your child has talked about suicide.

Is it right to monitor a teenager at all?

It depends entirely on how. Reading a teenager's private conversations is monitoring, and most families come to regret it. A consented check that reads nothing back to you, and only speaks up when somebody may be being hurt, is closer to a smoke alarm in a hallway. Agree it out loud, together.

More guides for parents