A Chatbot for Young Minds? Young People Set Their Own Terms

US · CanadaAmérica Latina

Weekly AI Watch

Every week, what changed and what it means for you. No hurry, no alarm.

Week of 23 October 2026EN | ESPast editions18,276 papers found · 17,064 read · 5,349 belonged · 37 stories passed every check

Observatorio de la
Inteligencia Artificial

Cada semana, qué cambió y qué significa para usted. Sin prisa y sin alarma.

Semana del 23 de octubre de 2026EN | ESEdiciones anteriores18,276 estudios encontrados · 17,064 leídos · 5,349 correspondían · 37 notas pasaron todas las comprobaciones

weeklyAI · Week of 23 October 2026weeklyAI · Semana del 23 de octubre de 2026

← Your child← Sus hijos

interviews · NPJ digital medicine · la publicación, 19 jun 2026 · gratis

Lo que los jóvenes pidieron a un chatbot de salud mental: que no reemplace a nadie

En talleres con 32 jóvenes en Australia dijeron qué esperan de estas herramientas. El estudio recoge sus opiniones, no lo que un chatbot logra en la vida real.

Versión breve · la versión detallada sigue, unos 6 min

Pregunte al chat de weeklyAI

Pregúnteme por este estudio: a quiénes se estudió, qué encontró y qué no dice.

Las conversaciones se guardan mientras exista weeklyAI, para mejorar la publicación. Se responde en el idioma en que usted escribe.

El estudio, de un vistazo
Quiénes: Jóvenes australianos que probaron un chatbot de salud mental llamado Mia
Cuántos: 32
Dónde: Australia
Cuándo: 2025
Tipo de estudio: interviews
Quién lo hizo: Universidad de Sydney, Australia
El límite que importa: Eran mayores de 18 años, en Australia, y vieron prototipos una sola vez, no un producto final.

Cuando vea una app o un chatbot de salud mental ofrecido a su familia, busque si dice con claridad para qué sirve, cómo decide lo que decide y en qué momento pasa la conversación a una persona.
Lectura de weeklyAI

Ver · 0:29

"It can't outright replace actual psychologists and psychiatrists, because the human connection is still important at the end of the day." Un joven lo dijo sin rodeos: la conexión humana sigue importando. La frase salió de uno de los talleres en línea del estudio, aprobado por la Universidad de Sydney, con 32 jóvenes.

Esa idea atravesó las conversaciones. Les mostraron Mia, un chatbot hecho primero para profesionales de servicios de salud mental juvenil, y les preguntaron qué debería ser para ellos y cómo debería encajar en los servicios. El estudio se hizo en Australia. De los 32 participantes, 19 respondieron una encuesta opcional y tenían entre 18 y 30 años.

Lo que pidieron fue concreto. Querían que el chatbot explicara cómo llega a sus conclusiones, mostrara en qué evidencia se basa y dejara claro qué hace con sus datos. También pidieron una división: sus conversaciones tal cual, privadas; los resúmenes que el sistema arma a partir de ellas, visibles para el personal clínico. Así, dijeron, se puede cuidar a alguien en riesgo sin exponer todo lo que escribió.

Pidieron además que note cuándo alguien está en peligro y lo dirija a profesionales. Y que ofrezca opciones: escribir o hablar, respuestas cortas o largas, una app o una página web. Los jóvenes rechazaron que un chatbot de este tipo funcione como amigo o compañero.

El estudio registra lo que estos jóvenes dicen que quieren. No prueba que Mia haya aumentado la confianza de nadie ni que haya mejorado la salud mental de alguien. Tampoco hubo comparación con otra herramienta. Los participantes vieron prototipos que funcionaban en un solo momento, no un producto terminado durante meses.

Queda otro límite. Eran mayores de edad, en Australia, y se ofrecieron a participar en una investigación sobre salud mental. Nada de esto viene de menores de 18 años, de los jóvenes más marginados ni de familias en América. Tómelo como lo que dicen adultos jóvenes, no como una guía probada para su hijo.

Aun así, sus peticiones señalan qué revisar. Cuando vea una app o un chatbot de salud mental ofrecido a su familia, busque si dice con claridad para qué sirve, cómo decide lo que decide y en qué momento pasa la conversación a una persona.

Qué significa para usted

Los propios jóvenes del estudio pidieron que estas herramientas expliquen cómo llegan a sus conclusiones y que pasen la conversación a una persona cuando alguien está en riesgo. Esa es la pregunta que usted puede llevar consigo cuando le ofrezcan un chatbot para su hijo. El estudio no midió si alguno de estos programas ayuda de verdad, y sus participantes eran mayores de edad y vivían en Australia.

Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9

Quién pagó: El trabajo fue financiado por un donante filantrópico anónimo afectado por la salud mental, con apoyo adicional de subvenciones del National Health and Medical Research Council a dos autores y una beca del Australian Government Research Training Program; el artículo no indica si los financiadores tuvieron alguna influencia, y no se atribuyó ningún equipo o software a terceros.

No tome esto como consejo médico profesional.

Versión detalladaLos pasajes copiados del artículo, las ilustraciones y cada fuente con cuánto leímos de ella · 1257 palabras · unos 6 minLeerla →Cerrar

Los jóvenes que probaron un chatbot de salud mental pidieron tres cosas antes de confiar en él

Un estudio australiano con 32 participantes de 18 a 30 años describe lo que esperan de estas herramientas. No mide si ayudan ni si dañan.

Un joven escribe en su celular que no ha dormido bien en semanas y que siente que nada le importa. Del otro lado no hay una persona: hay un programa entrenado con grandes cantidades de texto que responde, pregunta y acomoda la información en un resumen. El estudio que nos ocupa puso ese programa frente a 32 jóvenes australianos, de 18 a 30 años, en talleres en línea, y les preguntó qué pensaban de él¹.

El programa se llama Mia, Mental health Intelligence Agent. Nació como herramienta para profesionales de servicios juveniles en Australia y el estudio lo rediseñó para consumidores con la ayuda de esos jóvenes¹. Los investigadores analizaron las conversaciones de los talleres con un método cualitativo que busca temas compartidos, sin partir de categorías previas¹.

Del análisis salieron cuatro temas. Los jóvenes aceptaron el chatbot como apoyo, siempre que acompañe y no sustituya la relación humana en el cuidado². Lo dijeron con claridad: "No puede reemplazar de plano a psicólogos y psiquiatras reales, porque la conexión humana sigue importando; no es algo que la IA o los chatbots puedan replicar"³. La confianza, para ellos, depende de que el sistema sea transparente y certero⁴.

Piense en el mostrador de una farmacia: el frasco trae la lista de ingredientes, la dosis y las advertencias. Los participantes pidieron algo parecido. Quisieron poder revisar en qué evidencia se basa el chatbot, cómo llega a sus conclusiones y qué pasa con sus datos⁵. Lo curioso es que esa exigencia no es capricho: en salud mental un error pesa, y saber cómo funciona un sistema es parte de saber si se puede confiar en él⁴.

El mecanismo que describieron es sencillo de seguir. El chatbot recibe lo que escribe la persona, lo interpreta, detecta qué información falta y arma un resumen con recomendaciones, según la descripción del sistema en el apartado de métodos. Los jóvenes distinguieron dos capas: la conversación cruda, que debe quedarse privada, y los resúmenes que el sistema genera, que sí pueden llegar al clínico que los atiende⁶. Esa separación les permitía hablar sin filtro y, al mismo tiempo, dejar que alguien vigile el riesgo. Pasa lo mismo con la libreta de un terapeuta: el paciente escribe lo que quiere y el profesional lee lo que necesita para ayudar.

Sobre la seguridad, fueron directos. Un chatbot de salud mental debe reconocer cuándo alguien está en riesgo y responder de manera adecuada⁷.

¿Y qué lugar le toca a la herramienta? Los participantes la imaginaron como guía dentro de un servicio de salud, no como compañía. Distinguieron con cuidado ese papel del de amistad: un chatbot que funcione como amigo sería inapropiado en salud mental, donde los límites terapéuticos y la relación clínica son lo primero⁸. Ahí hay una concesión que conviene hacerles: un espacio amable y sin juicio puede animar a quien no se atreve a pedir ayuda a dar el primer paso. Pero el paso siguiente tiene que llevar a una persona.

El estudio no es el único que mira este terreno. Otra investigación siguió lo que miles de jóvenes estadounidenses escriben en aplicaciones de IA y encontró que el uso más frecuente es como herramienta, aunque también aparecen el juego de roles y el apoyo emocional, según el resumen de ese trabajo; pudimos leer solo el resumen⁹. Un estudio con 21 adolescentes que se autolesionan describió que recurren a tecnologías digitales, incluido ChatGPT, para regular emociones y hablar de temas delicados, y que valoran la inmediatez y la privacidad aunque notan respuestas genéricas o inexactas; también leímos solo el resumen¹⁰. Y una revisión de 95 estudios sobre aplicaciones de IA para adolescentes halló que casi todas reportan desempeño predictivo o diagnóstico, pero solo una midió efecto sobre síntomas; leímos solo el resumen¹¹.

Conviene decir con claridad qué no muestra este trabajo. Los participantes tenían entre 18 y 30 años y vivían en Australia, así que nada aquí viene de menores de edad ni de familias de América Latina o de América del Norte¹². Usaron prototipos en un solo momento, no un producto terminado durante meses, de modo que lo que esperaban puede diferir de lo que vivirían en uso prolongado¹³. El estudio recoge percepciones y recomendaciones; no mide si el chatbot mejora o empeora el ánimo de nadie.

Los autores declaran sus intereses. El trabajo lo pagó un donante filantrópico afectado por la salud mental que prefirió el anonimato¹⁴. Uno de los autores tiene una participación accionaria en una empresa de transformación digital de servicios de salud mental, y otra autora reporta honorarios y asesorías de farmacéuticas. Son datos que el lector merece tener sobre la mesa.

Así lo leemos nosotros. Cuando una tarea delicada pasa a un sistema automático, mucha gente supone que ese sistema es más parejo y confiable que una persona, porque no se cansa ni tiene favoritismos. Es una expectativa razonable, y a la vez peligrosa: un programa que nunca se distrae tampoco reconoce cuándo se equivocó. En casa, eso se traduce en algo concreto: si su hijo o hija usa una herramienta de IA para hablar de su ánimo, pregúntenle si le parece más justa que una persona y por qué. La respuesta dice más de lo que parece. Sabremos que nos equivocamos si los jóvenes desconfían del chatbot justamente por ser máquina, o si piden menos explicaciones a un sistema que a un profesional.

Hay otra manera de verlo, y tiene que ver con la coherencia. La confianza entre personas se sostiene cuando lo que alguien dice coincide con lo que hace; si las palabras y los actos se separan, se rompe aunque el mensaje suene amable. Con estas aplicaciones pasa igual: si la herramienta promete proteger los datos y los comparte sin avisar, o promete apoyo y no deriva a ayuda en una crisis, la confianza se cae y cuesta recuperarla. Antes de dejar que un adolescente use una app de apoyo emocional, revisen juntos qué promete sobre sus datos y qué pasa en una crisis, y comprueben si de verdad lo cumple. Si algo no cuadra, es señal para preguntar antes de seguir.

Y hay un tercer hilo, que los padres reconocerán. En la adolescencia se necesita un espacio sin mirada ajena; por eso una respuesta inmediata y sin juicio puede resultar atractiva, y conviene preguntar qué le gusta de esa conversación. Lo que no debe perderse es que siga habiendo adultos y profesionales con quienes hablar de lo que le pasa.

¿Qué tendría que pasar para que estas ideas se volvieran realidad? Los propios autores lo plantean: procesos de entrada transparentes, la herramienta presentada como complemento del cuidado humano, rutas de intervención clínica en crisis y reglas claras sobre los datos¹⁵. Nada de eso está probado aquí; son condiciones que los jóvenes pusieron sobre la mesa.

Volvamos al joven que escribe de noche en su celular. Lo que este estudio ofrece no es una respuesta sobre si el programa ayuda, sino una lista de lo que él y sus pares querrían ver antes de confiar: saber cómo funciona, saber quién lee qué y saber que, si algo se pone grave, hay una persona del otro lado. Esa lista es una buena pregunta para llevar a la próxima conversación sobre tecnología y salud mental en su casa: ¿quién revisa lo que este programa concluye sobre mí?

De dónde sale cada dato de contexto, y cuánto leímos de cada documento

Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "We examined the Mental health Intelligence Agent (Mia), a genAI chatbot originally designed for professionals in Australian youth services. Following co-design, 32 young people participated in online workshops exploring their perceptions of genAI chatbots in youth mental health"
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "GenAI chatbots are framed as supportive tools within youth mental health, provided they augment rather than replace the relational and empathetic role of human care."
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "It can’t outright replace actual psychologists and psychiatrists, because the human connection is still important at the end of the day, it’s not something that can be replicated by AI or chatbots"
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Trust in genAI chatbots in youth mental health depends on system transparency and accuracy."
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Young people expected transparency across multiple dimensions, including being able to inspect a genAI chatbot’s evidence base, examine its decision-making processes, and have clarity on how their data would be used and stored."
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "While raw conversational data should remain private, AI-generated insights developed from genAI chatbot interactions (i.e. summaries detailing a user’s needs, personalised treatment, self-management, and service navigation recommendations, and deterioration or risk detection) should be accessible to clinicians and services."
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "On interaction safety, young people emphasised that genAI chatbots should recognise and respond appropriately when a user is at risk."
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Importantly, young people distinguished the guide role from companionship or friendship, emphasising that a genAI chatbot functioning as a friend would be inappropriate in mental health contexts where therapeutic boundaries and clinical relationships are paramount."
Maheux AJ, Boeldt D, Akre-Bhide S, Flannery J, Paige A, Villella G, et al. (2026). How US Youth Use AI Chatbots: Conversation Patterns From Naturalistic Keystroke Observations. Journal of Medical Internet Research. 10.2196/95819 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Results identified 11 mutually inclusive themes, including functional tool use; social practice; role-play, including general, romantic, and sexual; emotional support; friend-like interactions; and violence."
Ramirez I, Barnas O, Mohr DC, Reddy M, Washburn JJ, Nguyen T, et al. (2026). Navigating self-injury in a digital world: adolescents’ perspectives on coping, help-seeking, and technology. Frontiers in Digital Health. 10.3389/fdgth.2026.1905943 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Participants reported using a range of tools, including music streaming platforms, journaling and mood-tracking apps, online communities, and AI systems such as ChatGPT to regulate emotions, reflect on their experiences, seek advice, and discuss sensitive topics."
Dhamayanti M, Susaldi S, Hanifa F, Agustini Y, Emilia O, Hardinsyah H, et al. (2026). Artificial intelligence-enabled mental health applications for adolescents: a scoping review of safety, effectiveness and clinical implementation. BMJ Open. 10.1136/bmjopen-2026-125586 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Predictive or diagnostic performance was reported in 83 studies (87.4%), but symptom-level effectiveness in only 1 (1.1%)."
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "the study participants consisted of young people aged 18-30 who were willing to participate in mental health services research in the Australian context"
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "participants engaged with functioning prototypes at one time point rather than final products over an extended period of user testing, and thus, user experiences may differ from anticipated ones"
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "This work was supported by a philanthropic funding donor affected by mental health who wishes to remain anonymous."
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Implementation requires transparent onboarding processes and positioning genAI chatbots as complementary to human-delivered care. And governance requires clinician intervention pathways during crises, transparency about data practices, multi-stakeholder governance frameworks, and mechanisms for sustained lived experience input."

No tome esto como consejo médico profesional.

interviews · NPJ digital medicine · the paper, 19 Jun 2026 · free

A Chatbot for Young Minds? Young People Set Their Own Terms

In online workshops in Australia, 32 people aged 18 to 30 said a mental health chatbot should support therapists, never replace them. The study records what they want, not what any chatbot does.

Short version · the longer version follows, about 9 min

Ask the weeklyAI chat

Ask me about this study: who was studied, what it found, and what it does not say.

Conversations are saved for as long as weeklyAI exists, to improve the publication. Answers come in the language you write in.

The study at a glance
Who: young people in online workshops about a mental health chatbot
How many: 32
Where: Australia
When: between March and August 2025
Kind of study: interviews
Who did it: University of Sydney researchers
The limit that matters: Only adults 18 to 30 in Australia; no one under 18; early prototypes at one time point.

The next time an app or a chatbot is offered to your family, look for one that says plainly what it is for, how it decides things, and when it hands a conversation over to a person.
weeklyAI's reading

Watch · 0:50

"It can't outright replace actual psychologists and psychiatrists, because the human connection is still important at the end of the day, it's not something that can be replicated by AI or chatbots," one young person in the study said.

The researchers found this view among the young people they spoke to. The University of Sydney researchers ran the workshops online. The 19 people who filled in a follow-up survey were 18 to 30 years old.

The researchers showed them Mia, a chatbot first built for professionals in Australian youth services. They asked what it should be for young people and how it should fit into mental health services.

The answers were specific. Young people wanted the chatbot to explain how it reached its conclusions and to show the evidence behind them. They wanted their raw conversations kept private, while clinicians and services could see summaries the system made from them. They wanted it to notice signs of risk and point people to professionals or crisis services. They wanted choices about how they use it, by text or voice, on a phone or a website.

What they said they want is not the same as what any chatbot has been shown to do. The study followed no one over time and compared nothing. It cannot show that a chatbot improves mental health, or harms it.

The people in the study were adults, and they were in Australia. Their answers say nothing about children under 18, or about families in the United States, Canada or Latin America. And they used early versions of Mia at a single point in time. A finished product, used over months, may feel different.

The researchers heard worries too: that a chatbot might get facts wrong, treat people with bias, or encourage unhealthy dependence. The study recorded those worries. It did not measure them.

The next time an app or a chatbot is offered to your family, look for one that says plainly what it is for, how it decides things, and when it hands a conversation over to a person.

What this means for you

The study gives you a checklist, not a verdict: what a chatbot is for, how it reaches its answers, and whether it steps aside for a person. Ask those questions of any app marketed to your family, and treat clear answers as the minimum. No one in this study was under 18, and nothing here shows what a chatbot does to a teenager over time.

Who paid: The work was supported by an anonymous philanthropic donor affected by mental health, with additional support from National Health and Medical Research Council grants to two authors and an Australian Government Research Training Program scholarship; the article does not say whether funders had any say, and no equipment or software was credited to an outside party.

Do not take this as professional medical advice.

The longer versionThe passages copied from the paper, the pictures, and every source with how much of it we read · 1717 words · about 9 minRead it →Close

Young People Told Researchers What a Mental Health Chatbot Would Have to Prove. Nobody Has Tested It Yet.

A study from Sydney asked 32 young people, ages 18 to 30, how a therapy chatbot should behave. They gave conditions, not approval.

The chatbot is called Mia. It was built for professionals in Australian youth mental health services, and the researchers wanted to know whether it could be remade for the young people themselves. So they brought in 32 of them, ages 18 to 30, for online workshops between March and August 2025. The workshops came in two kinds: five sessions to discuss and design, and four sessions where participants used a working prototype on their own while saying their thoughts aloud. The analysis produced four themes. Young people wanted artificial intelligence that sounds human without making care less human; they wanted to know what is under the hood; they asked whether a chatbot is the right tool, in the right place, at the right time; and they wanted to make it theirs on safe ground¹.

The first theme carries the clearest message. The young people treated a chatbot as a support, on one condition: that it add to human care rather than replace the relationship and empathy at its center². One participant put it plainly: a chatbot "can't outright replace actual psychologists and psychiatrists, because the human connection is still important at the end of the day"³. They also worried about what such a tool could get wrong. They raised made-up answers, poor recall of earlier conversations, bias in decisions, unhealthy dependence, and the risk of feeling judged. During testing, some tried to break the system on purpose, with vague or complicated descriptions of distress and with input in languages other than English.

Trust, in their account, rests on transparency and accuracy⁴. They wanted to inspect the evidence a chatbot draws on, follow its reasoning, and know how their data would be used and stored⁵. They also drew a line of their own between two kinds of information. The raw conversation should stay private, in their view. The summaries and recommendations built from it, including warnings that someone is deteriorating or at risk, should reach clinicians and services⁶. On safety they were direct: a chatbot should notice when a user is at risk and respond appropriately⁷. And they rejected one framing outright. A chatbot that acts as a friend is inappropriate in mental health care, they said, where therapeutic boundaries and clinical relationships come first⁸.

What the study cannot show matters as much as what it found. It measured opinions and design wishes, not outcomes. No one was compared with anyone, and no one's mental health was tracked. The participants were 18 to 30 and living in Australia, willing to take part in mental health services research⁹. Nobody under 18 was in the room, and the authors say so, along with the absence of voices most marginalized by stigma. The workshops ran remotely, and internet problems affected some transcripts. The participants used early prototypes at a single point in time, not a finished product over months, so what they expected may differ from what they would experience¹⁰. Only 19 of the 32 filled in the optional demographic survey, so the description of the group is partial. And the whole study concerns one chatbot in one country. The authors report that the tool was built on a multi-stage reasoning system with a knowledge bank drawn from the Youth Model of Care literature, clinical guidelines and expert annotations, and they describe results from a clinical evaluation against an expert panel. Those figures come from that separate evaluation, not from these workshops, and they say nothing about what the chatbot does for a young person.

Consider what a parent in Mexico City or Chicago would actually be handed if a service adopted a tool like this. The technology underneath is a large language model, the same family of system as the general chatbots already in wide use. It engages in conversation, sorts through large amounts of text, answers questions, translates between languages, and writes new content based on what the user types¹¹. That last part is the reason for both the promise and the worry: the system produces its answers rather than retrieving them from a fixed script. Young people are already taking general-purpose systems such as ChatGPT to mental health questions¹². A purpose-built tool would sit in that same habit, with clinical guardrails attached.

The scale of the need is not in dispute. Worldwide, young people aged 12 to 25 are experiencing high and, in most cases, rising rates of anxiety, depressive symptoms, psychological distress and other mood disorders¹³. The study's introduction lists rising rates of self-harm and suicide among young people; the workshops themselves did not examine anyone's care. Services are described as overburdened, with gaps in workforce and coordination. The appeal of a chatbot is easy to see: it answers at three in the morning, it does not keep a waiting list, and it never looks tired of you.

It is worth setting beside this study a different kind of look at the same question. Researchers monitored what young people actually typed into general AI apps on their phones, using a parental monitoring service installed by guardians. The study covered 3,363 young people in the United States aged 1 to 17, according to the summary; we could read only the summary, the full paper is behind a subscription¹⁴. Most use was practical: about 62 percent of user-app-days involved using the AI as a tool. But notable shares involved violence (15 percent), sexual role-play (15 percent), romantic role-play (8 percent), general role-play (9 percent) and friend-like interaction (6 percent)¹⁵. Violence and sexual or romantic role-play were relatively more common among younger users¹⁶. The authors note a limit of their own: only families who opted into the keyboard overlay and whose children used AI at all were included¹⁷. Read the two studies together and a gap appears. The Australian young people asked for a tool with boundaries and a clinician behind it. The American data show children wandering into role-play and companionship with general apps that have neither.

Another recent study interviewed 21 adolescents aged 14 to 18 with lived experience of self-injury, according to its summary; we could read only the summary¹⁸. These teenagers described using music apps, journaling tools, online communities and systems such as ChatGPT to steady themselves, reflect, get advice and talk about sensitive things¹⁹. They valued the accessibility, privacy and immediacy, and they named the limits too: answers that were inaccurate or generic, weak moderation in online groups²⁰. Stigma and worry about disclosure often kept them from seeking help in person²¹. The study's authors say the findings carry implications for clinicians and caregivers supporting adolescents²².

A wider review of AI-enabled mental health applications for adolescents found 95 primary studies, according to its abstract; we could read only the abstract²³. Most of the field is machine learning on structured data, and the leading function is predicting risk²⁴. Performance of some kind was reported in most studies, but symptom-level effectiveness in only one²⁵. No primary study provided a dedicated prospective safety evaluation²⁶, and none evaluated sustained routine-care scale-up or implementation cost as a principal outcome²⁷. The review's authors argue that future applications should be evaluated as supervised systems with a human in the loop and explicit safeguards for consent, privacy and fairness²⁸.

What would have to happen before any of this reaches a clinic your family uses? The study's own list is concrete. Onboarding would have to explain the tool's role and limits plainly, and the chatbot would have to be positioned as complementary to human-delivered care. Governance would need pathways for a clinician to step in during a crisis, transparency about data practices, frameworks involving more than one kind of stakeholder, and a standing way for lived experience to shape the system²⁹. None of that is a test result. It is a specification, written partly by the young people themselves.

Here is how we read it. The pattern in these workshops is a request for a witness, not a friend: someone or something that remembers what was said, notices when things are sliding, and hands the information to a person who can act. That is a modest and specific ask, and it is the kind of ask a family can carry into a conversation. If your teenager is already talking to a general chatbot about how they feel, expect that the warmth they describe may be the system mirroring their words back, and expect that limits on it will feel like a loss to them. You would know we are wrong if teenagers in this situation routinely push back on a chatbot's conclusions by themselves and tell an adult when something feels off. Until then, the practical move is small: agree at home that questioning the tool is expected, not rude, and ask now and then what the chatbot got wrong. Treat quiet agreement with it as something to check, not proof that it works.

We also read the American monitoring study as a reminder of where children actually are, not where a design document puts them. The young people in Sydney asked for a bounded guide with a clinician behind it; the keystroke data show children using general apps for role-play and company, with no clinician anywhere in the picture. That gap is not a reason to panic. It is a reason to ask a direct question of any service that offers your child an AI tool: who sees what it learns, and who is called when it notices something wrong? A service that cannot answer that plainly has not finished the work these 32 young people described.

The workshops opened with a prototype built for professionals and closed with a redesign: new screens explaining what the system does, what it does not do, and how much it knows about the user so far. That is what the young people asked for, and it is on the record. Whether the finished thing earns their trust is a question no workshop can answer. The next time a clinic or an app offers your child a chatbot for their mental health, the first question is now an easy one: who is behind it when it matters?

Where each piece of context comes from, and how much of it we read

Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "We examined the Mental health Intelligence Agent (Mia), a genAI chatbot originally designed for professionals in Australian youth services. Following co-design, 32 young people participated in online workshops exploring their perceptions of genAI chatbots in youth mental health"
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "GenAI chatbots are framed as supportive tools within youth mental health, provided they augment rather than replace the relational and empathetic role of human care."
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "It can’t outright replace actual psychologists and psychiatrists, because the human connection is still important at the end of the day, it’s not something that can be replicated by AI or chatbots"
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "Trust in genAI chatbots in youth mental health depends on system transparency and accuracy."
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "Young people expected transparency across multiple dimensions, including being able to inspect a genAI chatbot’s evidence base, examine its decision-making processes, and have clarity on how their data would be used and stored."
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "While raw conversational data should remain private, AI-generated insights developed from genAI chatbot interactions (i.e. summaries detailing a user’s needs, personalised treatment, self-management, and service navigation recommendations, and deterioration or risk detection) should be accessible to clinicians and services."
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "On interaction safety, young people emphasised that genAI chatbots should recognise and respond appropriately when a user is at risk."
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "Importantly, young people distinguished the guide role from companionship or friendship, emphasising that a genAI chatbot functioning as a friend would be inappropriate in mental health contexts where therapeutic boundaries and clinical relationships are paramount."
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "the study participants consisted of young people aged 18-30 who were willing to participate in mental health services research in the Australian context"
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "participants engaged with functioning prototypes at one time point rather than final products over an extended period of user testing, and thus, user experiences may differ from anticipated ones"
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "Advances in large language models (LLMs) and machine learning techniques present an opportunity for introducing conversational generative AI agents (or genAI chatbots) in youth mental health to engage users, classify large volumes of text, answer questions, translate languages, and generate new content based on user input"
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "This aligns with the existing trend of young people turning to publicly accessible, general-purpose systems, such as OpenAI’s ChatGPT, for mental health support"
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "Worldwide, young people aged 12-25 are experiencing high and, in most cases, rising rates of anxiety, depressive symptoms, psychological distress, self-harm, suicide, and other mood disorders"
Maheux AJ, Boeldt D, Akre-Bhide S, Flannery J, Paige A, Villella G, et al. (2026). How US Youth Use AI Chatbots: Conversation Patterns From Naturalistic Keystroke Observations. Journal of Medical Internet Research. 10.2196/95819 — only the abstract - the full text could not be fetched — the passage: "Participants were 3363 US youth aged 1-17 years using the Aura parental monitoring app, a commercially available service that guardians install on children's mobile devices for digital safety."
Maheux AJ, Boeldt D, Akre-Bhide S, Flannery J, Paige A, Villella G, et al. (2026). How US Youth Use AI Chatbots: Conversation Patterns From Naturalistic Keystroke Observations. Journal of Medical Internet Research. 10.2196/95819 — only the abstract - the full text could not be fetched — the passage: "Most user-app-days involved using AI as a tool (26,343/42,355, 62%), but notable proportions included violence (6460/42,355, 15%), sexual role-play (6393/42,355, 15%), romantic role-play (3538/42,355, 8%), general role-play (3707/42,355, 9%), and friend-like interaction (2624/42,355, 6%)."
Maheux AJ, Boeldt D, Akre-Bhide S, Flannery J, Paige A, Villella G, et al. (2026). How US Youth Use AI Chatbots: Conversation Patterns From Naturalistic Keystroke Observations. Journal of Medical Internet Research. 10.2196/95819 — only the abstract - the full text could not be fetched — the passage: "Theme prevalence varied by age: tool use was more common among older adolescents, whereas violence and sexual or romantic role-play were relatively more common among younger users."
Maheux AJ, Boeldt D, Akre-Bhide S, Flannery J, Paige A, Villella G, et al. (2026). How US Youth Use AI Chatbots: Conversation Patterns From Naturalistic Keystroke Observations. Journal of Medical Internet Research. 10.2196/95819 — only the abstract - the full text could not be fetched — the passage: "Participants included only those who opted into the overlay Aura keyboard (to collect keystroke data) and who engaged in some degree of GenAI use."
Ramirez I, Barnas O, Mohr DC, Reddy M, Washburn JJ, Nguyen T, et al. (2026). Navigating self-injury in a digital world: adolescents’ perspectives on coping, help-seeking, and technology. Frontiers in Digital Health. 10.3389/fdgth.2026.1905943 — only the abstract - the full text could not be fetched — the passage: "We conducted semi-structured interviews between May 2024 and April 2025 with 21 adolescents (14-18 years old) with lived experience of self-injury."
Ramirez I, Barnas O, Mohr DC, Reddy M, Washburn JJ, Nguyen T, et al. (2026). Navigating self-injury in a digital world: adolescents’ perspectives on coping, help-seeking, and technology. Frontiers in Digital Health. 10.3389/fdgth.2026.1905943 — only the abstract - the full text could not be fetched — the passage: "Participants reported using a range of tools, including music streaming platforms, journaling and mood-tracking apps, online communities, and AI systems such as ChatGPT to regulate emotions, reflect on their experiences, seek advice, and discuss sensitive topics."
Ramirez I, Barnas O, Mohr DC, Reddy M, Washburn JJ, Nguyen T, et al. (2026). Navigating self-injury in a digital world: adolescents’ perspectives on coping, help-seeking, and technology. Frontiers in Digital Health. 10.3389/fdgth.2026.1905943 — only the abstract - the full text could not be fetched — the passage: "Although these technologies were valued for their accessibility, privacy, and immediacy, participants also highlighted limitations, including inaccurate or generic AI responses, limited moderation in online communities, and the complex emotional impacts of music."
Ramirez I, Barnas O, Mohr DC, Reddy M, Washburn JJ, Nguyen T, et al. (2026). Navigating self-injury in a digital world: adolescents’ perspectives on coping, help-seeking, and technology. Frontiers in Digital Health. 10.3389/fdgth.2026.1905943 — only the abstract - the full text could not be fetched — the passage: "Stigma and concerns about disclosure often limited help-seeking and contributed to their use of digital technologies to access information and support."
Ramirez I, Barnas O, Mohr DC, Reddy M, Washburn JJ, Nguyen T, et al. (2026). Navigating self-injury in a digital world: adolescents’ perspectives on coping, help-seeking, and technology. Frontiers in Digital Health. 10.3389/fdgth.2026.1905943 — only the abstract - the full text could not be fetched — the passage: "These insights have important implications for the development of future interventions and for clinicians, caregivers, and other health professionals supporting adolescents' mental health."
Dhamayanti M, Susaldi S, Hanifa F, Agustini Y, Emilia O, Hardinsyah H, et al. (2026). Artificial intelligence-enabled mental health applications for adolescents: a scoping review of safety, effectiveness and clinical implementation. BMJ Open. 10.1136/bmjopen-2026-125586 — only the abstract - the full text could not be fetched — the passage: "The database searches identified 11 167 records, and 1 additional evidence synthesis was identified and assessed during revision; 103 sources were included, comprising 95 primary studies and 8 evidence syntheses."
Dhamayanti M, Susaldi S, Hanifa F, Agustini Y, Emilia O, Hardinsyah H, et al. (2026). Artificial intelligence-enabled mental health applications for adolescents: a scoping review of safety, effectiveness and clinical implementation. BMJ Open. 10.1136/bmjopen-2026-125586 — only the abstract - the full text could not be fetched — the passage: "Structured-data machine learning was the dominant modality (42/95, 44.2%), and risk prediction or prognosis was the leading clinical function (45/95, 47.4%)."
Dhamayanti M, Susaldi S, Hanifa F, Agustini Y, Emilia O, Hardinsyah H, et al. (2026). Artificial intelligence-enabled mental health applications for adolescents: a scoping review of safety, effectiveness and clinical implementation. BMJ Open. 10.1136/bmjopen-2026-125586 — only the abstract - the full text could not be fetched — the passage: "Predictive or diagnostic performance was reported in 83 studies (87.4%), but symptom-level effectiveness in only 1 (1.1%)."
Dhamayanti M, Susaldi S, Hanifa F, Agustini Y, Emilia O, Hardinsyah H, et al. (2026). Artificial intelligence-enabled mental health applications for adolescents: a scoping review of safety, effectiveness and clinical implementation. BMJ Open. 10.1136/bmjopen-2026-125586 — only the abstract - the full text could not be fetched — the passage: "No primary study provided a dedicated prospective AI-safety evaluation."
Dhamayanti M, Susaldi S, Hanifa F, Agustini Y, Emilia O, Hardinsyah H, et al. (2026). Artificial intelligence-enabled mental health applications for adolescents: a scoping review of safety, effectiveness and clinical implementation. BMJ Open. 10.1136/bmjopen-2026-125586 — only the abstract - the full text could not be fetched — the passage: "No study evaluated sustained routine-care scale-up or implementation cost as a principal outcome."
Dhamayanti M, Susaldi S, Hanifa F, Agustini Y, Emilia O, Hardinsyah H, et al. (2026). Artificial intelligence-enabled mental health applications for adolescents: a scoping review of safety, effectiveness and clinical implementation. BMJ Open. 10.1136/bmjopen-2026-125586 — only the abstract - the full text could not be fetched — the passage: "Future applications should be evaluated as supervised, human-in-the-loop systems with explicit safeguards for consent, privacy, fairness"
Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "Implementation requires transparent onboarding processes and positioning genAI chatbots as complementary to human-delivered care. And governance requires clinician intervention pathways during crises, transparency about data practices, multi-stakeholder governance frameworks, and mechanisms for sustained lived experience input."

Do not take this as professional medical advice.

The free weekly drip by email

One email a week: the free story of every section, with its picture, and the headlines of the rest.

No payment, nothing to cancel. Unsubscribe from any issue.

El goteo semanal gratis por correo

Un correo por semana: la nota gratuita de cada sección, con su imagen, y los titulares de las demás.

Sin pago y sin nada que cancelar. Puede darse de baja desde cualquier envío.