It's 2 am and your kid is burning up.

Not dangerously, but you're not sure. And at 2 am, not sure is a bad place to be.

A few years ago, you'd have called your mother or waited till morning and taken the kid to the doctor.

But now there's a third option and it’s always awake… ChatGPT.

We all have done it and turns out we're not the exception.

A 2026 report by Tonic Worldwide found that 52% of young Indian parents now go to AI for parenting advice.

I get it…

The old way meant 10 browser tabs, each one scarier than the last, none of them about your actual kid. AI gives you one calm answer that sounds like it's only talking to you.

Urvashi Aneja, who runs the Digital Futures Lab, told UNESCO that parents go to chatbots to avoid the exhaustion of endless online searching. 

Anyone raising a kid with the whole internet in their pocket knows it, but then the same calm voice might also tell you the worst thing.

You ask about the fever, and somewhere in that neat little answer, it mentions the rare serious cause. 

The one in a thousand. It's just being thorough. But now you're scared of something you hadn't even thought of. 

The same UNESCO report noticed this, how the tech brings up the scariest possibilities, like brain damage after a simple fall.

So you ask one more question and then probably one more. Just to be sure.

That one-more-question is one of the most studied things in anxiety. Psychologists call it reassurance-seeking.

And it’s something like this:

Asking is what keeps it alive!

The Anxiety and Depression Association of America is clear on this. 

Reassurance-seeking keeps the anxiety alive, because you never get to learn the one thing that would settle you, that you can sit with not knowing and still be fine. 

Clinicians even have a term for it. 

They call it a safety behaviour, the kind of thing that feels helpful but makes the problem worse.

Now think about what a chatbot is. Something that gives you reassurance at any hour, never gets tired of you, never tells you that's enough.

It might be the best reassurance machine ever built. Which, going by all this, makes it very good at feeding the exact worry it's calming.

Some parents don't just do this at 2 a.m. They do it for a year.

Sadia Fuzail is a copy editor in Delhi. Her son Ilhan is a high-energy kid. In 2024, when he was three, his grandparents started worrying his behaviour looked like autism or ADHD.

Then her friends mentioned it too. The first time she heard it, she asked a chatbot. It told her he was developing normally.

For the next year, that chatbot was her main advisor. Every time anyone brought it up, she had the AI's answer ready as a shield.

A real therapist finally saw Ilhan and confirmed the AI was right.

But think about that gamble. For a year, the biggest question about her son's early development had one advisor. A machine that had never met him. 

This time it was right. Next time it might not be. And a year is a long time to lose.

The advice has to come from somewhere and where it comes from isn't here.

Researchers have measured it. A 2024 study in PNAS Nexus tested five of the biggest AI models against the World Values Survey, the standard map of how different cultures think. 

The models all clustered around Western, English-speaking values, and drifted furthest from non-Western ones, especially when you prompt them in English. 

A separate 2025 review of 17 studies reached the same conclusion: these models default to Western norms even when you're talking to them in another language.

Fuzail noticed it without needing the study. The answers, she said, lean towards life in the West.

You catch it in small ways. 

The model has a confident answer for all of it, in the same steady voice, and that voice has no idea which country it's in. 

And confident, fluent advice is the hardest kind to question, because nothing in it tells you it wasn't built for you.

There's a small fix, by the way. The same researchers found you can pull the bias down a little by telling the model to answer as an Indian parent would. 

It helps. It doesn't solve it. Most people don't know to do it anyway.

Step back and look at who used to answer these questions, before any of this.

It was never Google. 

It was the grandmother, bua, the neighbour who raised four kids. The doctor your family had gone to for generations.

All of them got plenty wrong. But they knew your kid. They knew your family, your food, your particular kind of chaos. When they were wrong, they were wrong in a way that still fit where you live.

We've handed all of that to one voice that's awake at 2 am and knows none of it.

The parent pouring everything into one child is often the same parent up at 2 am asking a machine if they're getting it right. Same level of anxiety, just a bit different.

The instinct behind all of it is good. Nobody opens a chatbot at 2 am because they don't care. They do it because they care a lot, and because not knowing is unbearable.

But parenting was never about having the right answer every time. It's about being okay with not knowing and doing your best anyway. The thing we keep reaching for is making us worse at exactly that.

I'm not saying delete the app. I use these tools every day, and they're genuinely useful for a lot of this. 

Akhil Damodaran, an academic and AI founder in the same report, put the line well. He's fine with a psychologist using AI to do their work, not with a patient using it instead of the psychologist.

So let me ask you what I keep asking myself. Last time you were worried about your child, who did you check with first? And did it make you calmer, or…?

Until next week,
Ritesh

P.S. (Full honesty): I asked an AI for feedback on this very piece. It gave me a confident, well-structured answer. I ignored most of it :P