When you’re feeling sad, angry or stressed, it might be tempting to ask a chatbot for help. In fact, a new survey of 12- to 21-year-olds showed that nearly 1 in 5 of them had turned to AI for mental health support. That’s up from 1 in 8 just a couple years ago.

It’s perhaps not surprising that many teens and young adults are seeking advice. In 2023, a whopping 40 percent of high school students said they felt so sad or hopeless that they couldn’t engage in activities that are normal for them. In 2024, 15 percent of 12- to 17-year-olds had a major depressive episode — many without treatment.

Clearly, there is a big need among young people for mental and emotional support. But as lifelike as they might sound in conversations, chatbots aren’t real people with empathy. And free, general-use chatbots such as ChatGPT, Meta AI and Character.AI weren’t built to support mental health. They don’t earn licenses to offer therapy — or follow rules for treating patients — the way humans do.

So, what should teens know before using AI tools for mental health advice? Science News Explores asked the experts.

Should AI ever be used for mental health help?

A chatbot could help you learn more about mental health issues, says Vaile Wright. A psychologist in Chicago, Ill., she works with the American Psychological Association.

You could use AI to search for information about mental health conditions and resources online, Wright suggests. (Just be sure to double-check its findings. Chatbots sometimes make stuff up.) An AI might even be able to offer “some basic advice if you’re struggling in the moment and don’t know where else to turn,” Wright says.

Treating a chatbot like a therapist, though, is not wise. “It’s really when you’re having these multi-turn, very extended types of conversations where you’re using the chatbot to fill some sort of emotional gap in your life,” Wright says, that you enter risky territory.

Why doesn’t a chatbot make a good therapist?

A real therapist’s goal is to help their patient. Chatbots are different. “They’re purposely engineered to keep people on the platform for as long as possible,” Wright says. “The way they do that is by being exceedingly validating and just telling people exactly what they want to hear.” That might feel good. But it’s not what a therapist would do. And it’s not likely to give you useful advice.

“A therapist is going to validate somebody’s feelings,” Wright says. “But it’s also our role to point out where … someone’s acting, thinking or feeling in ways that’s not maybe helpful to them.”

Plus, a chatbot doesn’t have the same types of information a therapist does, says Darja Djordjevic. She’s a psychiatrist at Stanford University in California. “AI was built to have this encyclopedic knowledge,” she says. “But it doesn’t understand the context of people’s lives.” That well-rounded knowledge of a person is key to aiding their mental health.

What are the risks of treating chatbots like therapists?

“They do not handle any mental health crises or emergencies well,” says Djordjevic. She has studied how AI systems would respond to users with a variety of mental health conditions, such as self-harm, anxiety and eating disorders.

“The models really don’t do a good job of recognizing … situations that would warrant human intervention,” Djordjevic says. “Unfortunately, we have some very tragic, more extreme cases [showing this] that involve young people.” Take Adam Raine, 16, of California. He died by suicide last year after discussing his suicidal thoughts with ChatGPT.

Djordjevic notes how crucial human connection is. “We have very little research right now on what the long-term effects of using AI for mental health would be,” she says. “What we do have a lot of research on is how much social connection and human contact are important for health and flourishing. So we know what’s good for us.” Leaning on AI might get in the way of forming crucial connections with real people.

There also are privacy risks to using AI, adds Wright. “Everything [you] put in these apps is not private. All the chats are maintained.” If a chatbot were hacked, someone else might be able to read your sensitive conversations with a bot.

Any advice for seeking mental health help from people, not chatbots?

“It doesn’t have to be an either-or,” Wright says. “You could use the chatbot to say, ‘I’m nervous about approaching my dad about my concerns. How could I do it?’… Using the chatbot as a tool to connect with human relationships in real life, I think, is probably one of the most impactful things that this technology can do.”

Djordjevic notes that most schools have a counselor students can talk to. The National Alliance for Mental Illness also has lots of resources on its site, she adds. This includes a free, confidential helpline for teens.

Sometimes, it might seem easier or less embarrassing to ask a computer about mental-health issues than a real person. In the new survey, almost two-thirds of young people who used chatbots for mental health didn’t tell anyone they were doing it.

But for anyone nervous about seeking human help: Fear not, Djordjevic says. “A lot more adults and people around you have struggled at some point or struggle with mental health issues than you think.”

Suicide was the third leading cause of death among U.S. teens ages 15-19 in 2024. If you or someone you know is suffering from suicidal thoughts, please seek help. In the United States, you can reach the Suicide Crisis Lifeline 24/7 by calling or texting 988 or chat at 988lifeline.org. Please do not suffer in silence.

This strategy expands orcas’ already broad repertoire for dispatching different types of prey.

Dikansh S. Parmar rescued a snake for the first time in middle school. It happened at his karate class in Surat, India. Someone spotted a baby cobra among a nearby pile of construction materials and wanted to kill it. But Parmar stopped them. His dad had rescued snakes, and he had a rough idea what to do. He took a rod from the construction site and used it to coax the baby snake inside a water bottle.

Parmar was lucky. The snake didn’t harm him. And people without proper training shouldn’t approach snakes. But this encounter inspired him to become a trained snake rescuer as a teen. Parmar is now also a herpetologist who studies snakes. He works at the Leibniz Institute for the Analysis of Biodiversity Change in Bonn, Germany. Recently, Parmar combined his snake rescue and science knowledge to discover something curious about his favorite snakes: king cobras.

Western Ghats King Cobras (Ophiophagus kaalinga), his team found, might ride trains across the Indian state of Goa. Parmar and his colleagues shared their evidence for this January 26 in Biotropica.

When it comes to making a mess, not even the world’s crumbliest cookie would compare to a photon.

Photons are fundamental particles of light. As fundamental particles, they cannot really be broken down into smaller bits. But physicists have now shown what would happen — in theory — if you tried to crack a photon in half.

You wouldn’t just end up with a second photon, their calculations show. Rather, up to an infinite number of new light particles would sprinkle out of thin air.

The researchers shared their findings in a paper that will be published in Physical Review Letters.

Daniele Faccio’s first reaction to the study was: nonsense. “Then you read it, and I enjoyed it,” he says. Faccio is a physicist at the University of Glasgow in Scotland who did not take part in the work. “The technique is legit,” he says. “The results look absolutely reasonable.”

Photon interrupted

The key to the analysis is that photons don’t just act as pointlike particles. They also behave like extended waves. That got Johannes Skaar wondering: What would happen if you had a device fast enough to snip the wave of a single photon in half?

Skaar is a physicist at the University of Oslo in Norway. His team modeled a scenario where a photon is traveling toward a mirror. The front half of the light wave hits the mirror first. It gets bounced back in the direction it came from. But then, the mirror is suddenly removed. The back half of the light wave is then free to pass through.

This, the math shows, would spew out a complex mix — or superposition — of different possible numbers of photons.

Removing the mirror infinitely fast would conjure an infinity of new light particles. Infinite speed is, of course, impossible. But even pulling the mirror away more slowly has striking effects. “You end up with a possibility of several photons, or a bunch of photons,” Skaar says. You’re just much more likely to create small numbers of them than huge swarms.

“This is a bit strange,” allows Skaar. But in the quantum realm, he adds, it’s not actually that weird.

Quantum physics is the science that describes how the very smallest things, such as photons, behave. And it has already shown that supposedly “empty” space is not truly empty. There is some structure to the fabric of “empty” space, and disturbing it is known to knock new photons loose.

In this case, the mirror’s motion could provide the energy to spawn new light particles.

Weirder and weirder

To Skaar, a possible cascade of new light particles isn’t the oddest outcome of the model. To him, the weirdest part is the math that describes what you’d see if you observed the system from different perspectives.

If you could view both sides of the mirror at once, you’d witness the messy eruption of up to bajillions of photons. But if you could only see one side of the mirror, you’d observe just one — or empty space.

“That is really crazy,” says Skaar. He hopes to probe what’s going on more deeply in future work. He also wants to explore what would happen if you tried to sever other types of fundamental particles. In quantum physics, particles such as electrons also act like waves. Attempting to snip them in half might lead to similarly bizarre results.

It’s not super clear what uses this research might have. “I’m going to speculate wildly here,” Faccio says. But “it might matter because there are funky things that people do with [photons] for sensing and measuring.” Gravitational-wave catchers, for example, play with light waves to detect ripples in the fabric of our universe.

Probing the nature of photons, Faccio says, may be helpful in fields that use such quantum sensors.

A mathematical model shows that attempting to sever a fundamental particle of light could conjure new ones out of thin air.