Your AI chatbot habit, sad brains, and sound-wave therapy

Using AI as a friend, not a tool, is linked to worse mental health

Talking to an AI chatbot the way you'd talk to a therapist is linked to depression and social anxiety — but using it to get work done is not.

Think of two people with a smartphone calculator. One uses it to crunch numbers faster. The other has started asking it how they feel about their day and whether anyone really loves them. A team of researchers in China scanned the brains of 222 university students with high-resolution MRI and also asked them how they use AI tools — functionally (to study, write, or search) versus socio-emotionally (to talk, seek comfort, or manage feelings). The split in outcomes was sharp. Students who used AI functionally tended to have better grades and slightly larger volumes in the dorsolateral prefrontal cortex — the region behind your forehead that handles planning and self-regulation — as well as more efficient connections in the hippocampal network, which handles memory. Students who used AI socio-emotionally showed the opposite: higher rates of depression and social anxiety, and lower gray matter volume in areas linked to social processing, including the amygdala. Here's the important catch. This is a cross-sectional study — one snapshot in time — which means we cannot tell which came first. It's entirely possible that people who are already anxious or depressed turn to AI for emotional support, rather than the reverse. The brain differences might reflect pre-existing states, not consequences of AI use. The sample is also 222 Chinese university students — a narrow slice of humanity. And notably, only 6.8% of students reported frequent socio-emotional AI use; 82.5% used it mainly for functional tasks. So this is not a majority phenomenon. What the study adds is a first serious attempt to map AI behaviour onto brain structure. That's worth knowing.

Focused sound waves could reach deep brain targets without surgery

A beam of sound, aimed with millimetre precision, can nudge neurons deep inside your brain — no scalp electrodes, no implants, no scalpel.

If you've ever seen a surgeon use ultrasound to break up a kidney stone without cutting the body open, you have the right mental image. Transcranial focused ultrasound — tFUS for short — works on a similar principle, but tuned down to gentle nudges rather than destruction. A review paper synthesising existing research describes how tFUS can reach brain structures like the amygdala, hippocampus, and thalamus — all deep inside the skull — and modulate their activity with millimetre-level precision. Compare that to standard techniques like transcranial magnetic stimulation, which affects centimetre-wide patches and cannot reach deep targets at all. Why does this matter for mental health? Treatment-resistant depression — the kind that doesn't respond to antidepressants — often involves dysfunction in exactly those deep structures. Current options for reaching them include brain surgery or electroconvulsive therapy. A non-invasive, precise alternative would be a meaningful step forward. The review also describes closed-loop systems: devices that monitor brain electrical activity in real time and adjust the ultrasound beam on the fly, like a thermostat for neural circuits. Important caveats. This is a review article, not a clinical trial. The author is synthesising other people's experiments, not reporting new results. Most closed-loop tFUS work has been done in animals or small human proof-of-concept studies. We are years away from knowing whether this treats depression, anxiety, or anything else reliably. The beam coupling — keeping the transducer in stable contact with the skull — remains an engineering problem, though bioadhesive gels have shown promise for up to 35 days. Promising direction. Early days.

Sad brain states are more 'stuck' than happy ones — EEG confirms it

When your brain processes sadness, it settles into a deeper, more stable groove than when it processes happiness — and for the first time, researchers have measured that groove.

Imagine rolling a marble across a slightly bumpy surface. Some dips are shallow — the marble rolls through them easily. Others are deep — the marble gets stuck. A team of researchers measured something analogous in twenty healthy adults using EEG, the technique that records the brain's electrical signals through electrodes on the scalp. Participants looked at happy and sad faces while their brainwaves were recorded. The researchers then applied a mathematical tool called Hopfield energy — borrowed from physics — to measure how 'settled' the brain's network state was during each type of face. Sad face processing consistently produced lower, more negative energy values, particularly in the alpha brainwave band (roughly 8–13 Hz). In physics terms, a lower energy state is a more stable one — harder to leave. The alpha band effect was substantial, with a Cohen's d of 0.83, which in psychology counts as a large effect size. They also found that the more efficiently connected the brain's network was, the more negative its energy — meaning hyperconnected networks get more stuck. This is potentially important for understanding depression, where people classically report being unable to stop thinking sad thoughts. A more energetically stable sad state would be consistent with that clinical picture. But the catch is real: only twenty participants, all young healthy adults, all looking at standardised face images in a controlled lab. The gap between 'processing a sad face for 500 milliseconds' and 'being clinically depressed for months' is enormous. This is a measurement proof-of-concept, not a diagnostic tool. Worth watching if the same pattern holds in larger, clinical samples.

Look at these three stories together and you notice a single thread: mental health research is slowly gaining the ability to measure things it used to only describe. We can now map how brain structure correlates with digital behaviour patterns. We can model how different emotional states produce quantifiably different levels of neural stability. We can aim sound waves at centimetre-deep brain targets with millimetre accuracy. None of these tools is clinical yet — each one sits somewhere between 'interesting lab result' and 'useful treatment.' But the direction matters. For decades, psychiatry has relied heavily on self-report: how do you feel on a scale of one to ten? The field is building the instruments to ask the brain directly. The question is whether those measurements will eventually connect to things that help people, or whether they will remain elegant maps of a territory no one knows how to navigate.

The AI-and-brain study from China is the kind of work that needs replication with a more diverse sample — watch for follow-up studies in European or North American cohorts, ideally longitudinal ones that track the same people over time. On the tFUS front, several clinical trials targeting depression with focused ultrasound are underway; results from the University of Michigan's LIFT-TRD trial would be worth flagging when they publish. And if the Hopfield energy approach to emotional brain states gets applied to a clinical depression cohort, that paper will be the one to read.

• Nature News: What does AI do to the teenage brain?
• Wikipedia: Transcranial focused ultrasound
• Quanta Magazine: The brain's electrical rhythms, explained