In the first article in this series I showed you how to ask AI about your child's symptoms well. This one is about what to do with the answer - because the most dangerous thing about AI health advice is not that it is wrong. It is that it is wrong in exactly the same confident, well-organised voice it uses when it is right.

The four ways AI answers mislead

1. Confident tone, wrong content

AI tools do not hedge by accident; smooth language is what they are built to produce. A made-up fact and a solid one arrive formatted identically. Your instinct that "it sounds sure, so it must know" is the exact instinct to switch off. In medicine, certainty has to be earned from evidence - not from sentence quality.

2. Invented sources

Ask an AI for references and it may produce a perfect-looking citation - journal, year, authors - that does not exist. This happens often enough to have a name among people who build these tools. If a study matters to a decision about your child, search for its title yourself. Thirty seconds. If you cannot find it, treat the claim attached to it as unproven.

3. Missing the thing that makes your child different

Even a correct general answer can be wrong for your child - because of age, a medicine they already take, an allergy, a condition you did not think to mention. General knowledge is not personal medicine. This is why the five context details matter so much, and why some questions should not go to AI at all (see the bright lines).

4. Stale medicine

Guidelines change - vaccine schedules, fever advice, dosing recommendations all get updated. An AI's knowledge has a cutoff date it rarely volunteers, and it may state an old recommendation as current fact. Anything involving doses, schedules or "what is recommended now" deserves a check against a live source.

The two-minute sanity check

Before an AI answer changes what you do - or ruins your evening - run it through this:

  • Would this change my actions? If the answer is just explanation, read on. If it would make you start, stop or skip anything medical, verify first. The bigger the action, the stronger the verification it needs.
  • Does an official source agree? Check the claim against the American Academy of Pediatrics (healthychildren.org) or the Indian Academy of Pediatrics (iapindia.org). Two minutes, one search. If official guidance says something different, believe the guideline.
  • Is it using fear? Answers - and the posts AI learned from - that push you toward panic, miracle cures or "doctors won't tell you this" are marketing wearing a stethoscope. Real medicine is calmer and less exciting than that.
  • Does it fit what I see? You are looking at the actual child. If the answer predicts a child in crisis and yours is drinking juice and negotiating for a cartoon, trust your eyes and get a human opinion rather than an escalated panic.
  • Is this one of the three bright lines? Doses, diagnoses, emergencies - if the question crosses one, close the app and use the phone for a phone call instead.

The rule of thumb I use myself

Treat AI like a well-read friend, not a doctor. Wonderful for understanding, explaining, preparing your questions, and knowing what the red flags even are. Not licensed to decide. When the stakes rise - a medicine, a decision to wait, a symptom that is getting worse - that is where the friend hands you over to a person who can actually examine your child.

Used in that order - AI to understand, sources to verify, a doctor to decide - you get the best of all three. That is the whole system. It is not anti-technology. It is just medicine, with the tools in their correct places.