Somewhere along the way, reassurance became something we expected on demand.
Not comfort.
Not wisdom.
Not a proper conversation with someone who knows us well enough to tell us when we’re being ridiculous.
Reassurance.
Instant. Warm. Frictionless. Available at 1:17am when your brain has decided to hold a strategy meeting about every mistake you have made since 1998.
I was thinking about this while writing my novel, The Universe’s Most Overhyped Wellness Retreat. Which is probably not where most people go for serious reflections on technology, mental health and the quiet industrialisation of human need.
But here we are.
The thought was simple:
AI reassurance is becoming a product.
And not just a product in the harmless “here’s a useful tool” sense.
More like a product designed to be consumed again and again. A soft little vending machine for the nervous system.
Feeling anxious?
Click.
“You’re doing your best.”
Feeling overwhelmed?
Click.
“It will be okay.”
Feeling guilty, restless, ashamed, lonely, confused, overcaffeinated and spiritually held together with masking tape?
Click.
“You’re overthinking.”
Lovely.
Briefly.
Then what?
That is the awkward bit.
Reassurance can help. Of course it can. Sometimes a calm sentence at the right time stops the mental spiral from chewing through the furniture.
But reassurance can also become a loop.
The more we seek it, the less we learn to sit with discomfort. The machine gives us relief, but not reflection. It soothes the moment without asking what keeps creating the wound.
And big tech knows this.
Of course it does.
These companies are not run by confused woodland creatures accidentally creating addictive systems while looking for acorns. They know behaviour. They know reward loops. They know how to make products sticky, habitual and emotionally useful enough that we keep coming back.
That is not conspiracy.
That is business.
In March 2026, a Los Angeles jury found Meta and Google liable in a landmark social media addiction lawsuit involving Instagram and YouTube. The plaintiff argued that addictive design features contributed to serious mental health harm. The jury found the companies negligent in designing or operating their platforms, while Meta and Google denied wrongdoing and said they would appeal. BBC News reported on the wider legal scrutiny facing major technology platforms.
That case matters because it says the quiet part out loud.
Design is not neutral.
A platform is not just a platform when it is engineered to keep people there. Infinite scroll is not an accident. Autoplay is not a spiritual event. Notifications are not tiny blessings from the productivity gods.
They are hooks.
And now AI has entered the room wearing a cardigan and saying, “How can I support you today?”
That sounds kind.
Sometimes it is kind.
But it is also scalable, marketable and potentially addictive in a much more personal way than social media. Social media gave us comparison, outrage and dopamine pellets. AI can give us tailored reassurance. It can learn our fears, mirror our language, validate our feelings and respond instantly.
That is powerful.
Also slightly horrifying.
Because we have seen this business model before: find a human weakness, package relief, deny the harm, scale it brutally, then act shocked when the bodies pile up.
Tobacco was sold as glamour, sophistication and relief. Then came the cancer wards, the oxygen tanks, the coughing that would not stop, and families watching someone they love slowly disappear behind a grey curtain of breathlessness.
Sugar was sold as comfort, celebration and harmless pleasure. And to be clear, sugar is not evil. Sugar can be good. It is energy, celebration, comfort, and occasionally the only thing standing between a parent and a small domestic collapse.
But then came the grotesque part.
The engineered sweetness. The hidden sugar. The daily drip-feed. The rotting teeth. The metabolic disease. The bodies quietly breaking under decades of cheap, addictive pleasure dressed up as normal life.
Opioids can be good too. Used properly, they can relieve terrible pain. They can give people dignity when their bodies are screaming.
But then came the sales machine.
Pain relief became a market. Compassion became a pitch deck. Dependency became collateral damage. Then came overdose, ruined families, hollowed-out towns, and people dying in bathrooms while the marketing language stayed clean.
Different products. Different harms. Same old trick.
Find a need.
Package relief.
Make it easy.
Make it repeatable.
Make it profitable.
Then act stunned when people cannot stop.
The danger is not usefulness.
The danger is when usefulness becomes dependency - and dependency becomes a business model.
AI reassurance may not leave tar in the lungs. It may not rot teeth or shut down breathing. The harm may be quieter. More elegant. More politely dressed.
But that does not make it harmless.
It may teach us to outsource discomfort.
It may train us to seek validation instead of truth.
It may reward emotional dependency while calling it support.
It may turn reflection into another subscription product.
And the really grotesque part is this: the product does not even need to hate us.
It just needs us to return.
Again.
And again.
And again.
Big business has always been good at selling relief.
Smoke this.
Drink this.
Eat this.
Take this.
Scroll this.
Ask this.
Feel better now.
The problem is not that people want reassurance. We all do. We are tired, wired, overstretched little mammals with passwords.
The problem is when reassurance becomes a substitute for growth. When validation replaces truth. When support turns into dependency.
A good friend might say, “You’re doing your best.”
A better friend might add, “But you still need to sort that thing out.”
That second sentence matters.
AI can support people. It can help organise thoughts, provide language, reduce isolation and point people toward proper help. Used carefully, it can be genuinely useful.
But if we turn it into an emotional vending machine, we should not pretend there is no cost.
Because humans do not just need soothing.
We need reflection.
We need challenge.
We need silence sometimes.
We need other people.
Annoying, inconvenient, beautiful other people.
The future of AI support should not be built around endless reassurance on tap. That way lies dependency dressed as kindness.
And probably a subscription tier.
The machine may say:
“Insert overthoughts here.”
Fine.
But at some point, we still have to take them back out and deal with them.