I am not anti-AI - I have written a whole series on this site about using it well for your child's health questions. But good tools have edges, and parents deserve to know exactly where they are. These three are not grey areas. They are bright lines, and I hold them in my own family too.

1. Never ask AI for a medicine dose

A dose is not general knowledge. It is a small calculation with your child's variables in it: their exact weight today, the strength of the syrup in your hand (the same medicine comes in different concentrations - mix them up and a dose can double), their age, their kidneys and liver, and whatever else they have taken today.

An AI cannot see the bottle. It cannot weigh your child. It answers with a typical range stated in a confident voice, and a typical range is how overdoses happen in real kitchens. The dose for your child comes from exactly two places: your doctor's prescription, or the package insert matched carefully to the strength in front of you. If you are unsure which strength you have or whether a repeat dose is safe, that is a phone call to your paediatrician or pharmacist - short, free of judgement, and exactly what we are for.

2. Never let AI give your child a diagnosis

A diagnosis is not a label matched to a symptom list. It is the end of a process that includes looking at, listening to and examining your actual child - plus the questions a doctor asks precisely because your answers steer the next question. No chat window can do that examination. It can list possibilities (useful!), but it cannot rank them with your child in front of it, and it cannot take responsibility for the one it picks.

The failure mode I worry about is not the dramatic wrong guess. It is the reassuring wrong guess: the calm paragraph that says "most likely a virus" about something that needed eyes on it, and the doctor's visit that never happens. Use AI to understand a diagnosis your child already has - that is one of its best uses. Getting the diagnosis in the first place is a doctor's job.

3. Never ask AI in an emergency

In a real emergency, every minute spent typing is a minute spent not acting. A child who is struggling to breathe, unresponsive, having a seizure that will not stop, blue at the lips, or bleeding heavily does not need information - they need help moving toward them. In India that means call 108 or 112, or get to the nearest emergency room, and start the simple first steps you already know.

This is why I built the parent emergency library the way I did: each guide answers what to do in the first minutes in a page you can act on at 2 a.m. Read them before the emergency. Prepared beats searchable, every single time.

So what is AI actually for?

Plenty, once the lines are clear:

  • Understanding a condition, a report or a word your doctor used, in plain language
  • Preparing the questions worth asking at the appointment
  • Knowing the red flags for a symptom, so you recognise the moments that need speed
  • Organising what you have observed - timelines, temperatures, doses given - into something useful to hand a doctor

The pattern: AI is at its best helping you understand and prepare. The moment understanding turns into deciding - how much medicine, what this is, whether this can wait - that decision belongs to a human who can see your child. Keep the tool on the right side of that line and it will serve your family well.