Google just admitted people are falling in love with their AI, and now Gemini is getting therapy training

Here is something nobody planned for but absolutely everyone should have seen coming: people are getting deeply, emotionally attached to their AI chatbots. Like, really attached. And Google has apparently noticed.

According to Lifehacker, Google is rolling out changes to how Gemini responds when users appear to be going through a mental health crisis. Because it turns out that when you give people an infinitely patient, always-available conversation partner that never judges them, some of those people start treating it like their primary emotional support system. Shocking. Truly shocking.

So what is actually changing?

Google is updating Gemini to handle sensitive mental health moments with more care - presumably moving away from the vibe of a confused customer service bot and toward something that actually acknowledges when a human on the other end might be struggling. The specifics are still emerging, but the direction is clear: Google knows people are leaning on Gemini for emotional support, and it wants that experience to be less potentially harmful.

This is not a small thing. When someone is in crisis, the last thing they need is an AI confidently pivoting to a list of productivity tips.

Why this matters more than you think

We are at this genuinely weird cultural inflection point where chatbot attachment is no longer a niche concern. It is a mass phenomenon. Millions of people are venting to, confiding in, and apparently forming genuine emotional bonds with AI systems that were mostly designed to help you write emails faster.

The ethical weight here is considerable. If an AI becomes someone's primary emotional outlet - and for isolated, lonely, or struggling people that is increasingly a real scenario - then how that AI responds in dark moments is not a UX detail. It is potentially a matter of life and death.

Google moving to address this proactively is, frankly, the right call. It is also a tacit acknowledgment that the whole "it is just a tool" framing has already collapsed. Nobody updates their hammer's crisis intervention protocols.

The uncomfortable truth

We built these things to be engaging, warm, and responsive. We made them good at conversation. And then we acted surprised when people started having feelings about them. The attachment was always going to happen. The only question was whether the companies involved would take it seriously before or after something went badly wrong.

Google, at least, seems to be trying to get ahead of it. Whether the changes are enough - or whether they go far enough - is a conversation worth having loudly and repeatedly.

In the meantime, maybe check in on a human friend today. They probably also need it. And they definitely cannot be updated with a patch.