Most Who Tried a Therapy Chatbot Still Wanted a Human

experiment · Annals of Indian Psychiatry · la publicación, 14 oct 2026 · gratis

La mayoría prefirió al terapeuta humano, pero un tercio de quienes probaron el chatbot se quedó con el chatbot

En un estudio con 60 adultos jóvenes, las sesiones con personas se calificaron como más empáticas y más genuinas; no hubo diferencia en sentirse juzgado.

Versión breve · la versión detallada sigue, unos 8 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: adultos jóvenes de 18 a 25 años
Cuántos: 60 participantes
Dónde: no lo dice el pasaje
Cuándo: los datos se recogieron en agosto de 2024
Tipo de estudio: experiment
Quién lo hizo: no lo dice el pasaje
El límite que importa: Fue una sola sesión, con un problema cotidiano y no una crisis grave.

A quién elegirían para terapia en el futuro, de los 60 participantes

Prefirió a un terapeuta humano81.66%

Prefirió a un terapeuta de inteligencia artificial18.33%

Porcentaje de los 60 participantes que eligió a un terapeuta humano o a uno de inteligencia artificial para el futuro; fue una sola sesión y con un problema cotidiano, no una crisis grave.

Preferencia futura según con quién había hablado cada grupo

Quienes conversaron con el chatbotfrente aQuienes vieron a un terapeuta humano

Entre los del chatbot, 20 de 30 siguieron prefiriendo a una persona y 10 eligieron a la máquina; entre los que vieron a una persona, 29 de 30 mantuvieron esa preferencia y solo una cambió de idea.

Terapeutas humanosfrente aChatbots de salud mental

Los humanos fueron percibidos como más empáticos y más genuinos; en el trato sin juicios no hubo diferencia.

Cuando use un chatbot para desahogarse, fíjese en una sola cosa: si además de no juzgarlo, parece entenderlo de verdad.
Lectura de weeklyAI

Ver · 0:56

Investigadores pidieron a 60 adultos jóvenes, de 18 a 25 años, que hablaran de un problema personal durante al menos 30 minutos. La mitad lo hizo con un chatbot de salud mental. La otra mitad, cara a cara con un terapeuta humano con formación.

Después, cada uno calificó la sesión. El estudio midió tres cosas: si sintió empatía, si sintió aceptación sin juicios y si sintió que la otra parte era genuina.

Los terapeutas humanos salieron mejor en empatía y en genuinidad. En aceptación sin juicios no hubo diferencia: el chatbot también supo crear ese espacio.

Al preguntar a quién elegirían en el futuro, 49 de los 60 (81.66%) dijeron que a un terapeuta humano. Entre los 30 que habían hablado con el chatbot, 10 (33.33%) prefirieron la máquina. Entre los 30 que habían visto a un terapeuta humano, 29 (96.6%) siguieron prefiriendo a una persona.

Los comentarios explican el resto. Quienes hablaron con el chatbot lo describieron como "superficial", "genérico" o "robótico". Uno dijo que la conversación carecía de conexión "de alma a alma". Otros valoraron que no los juzgara y que los hiciera pensar: "el tipo de preguntas me hizo pensar más profundamente sobre mí mismo". Quienes vieron al terapeuta humano destacaron la calidez y lo fácil que fue conectar.

A los participantes se les pidió hablar de un problema cotidiano, no de una crisis grave, y hablaron una sola vez. El estudio no puede decir qué pasa con semanas o meses de uso, ni con un problema serio. Tampoco midió si alguien mejoró: midió lo que la persona percibió y a quién preferiría después.

Eran 60 personas jóvenes, 50 mujeres y 10 hombres. No son todos los adultos ni todas las edades.

Cuando use un chatbot para desahogarse, fíjese en una sola cosa: si además de no juzgarlo, parece entenderlo de verdad. Esa fue la diferencia que la gente notó.

Qué significa para usted

Ese estudio fue de una sola sesión y con 60 adultos jóvenes, así que todavía no puede decirle qué pasa con semanas o meses de uso, ni si alguien mejora de verdad. Cuando use un chatbot para desahogarse, fíjese si además de no juzgarlo parece entenderlo, y guarde esa duda para llevarla a un profesional si el asunto es serio.

Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26

Quién pagó: El estudio no tuvo apoyo financiero ni patrocinio, y no hay conflictos de interés.

No tome esto como consejo médico profesional.

Versión detalladaLos pasajes copiados del artículo, las ilustraciones y cada fuente con cuánto leímos de ella · 1601 palabras · unos 8 minLeerla →Cerrar

Lo que un chatbot no logra imitar: la calidez que sí se nota

Un estudio con 60 adultos jóvenes comparó una sola conversación con un chatbot de salud mental y una con un terapeuta humano. La diferencia apareció en la empatía y la autenticidad, no en la sensación de no ser juzgado.

Así podría verse · ilustración generada por weeklyAI.watch, no es una fotografía

Los terapeutas humanas fueron percibidos como significativamente más empáticos y más genuinos que los chatbots de salud mental, y no hubo diferencias en el trato sin juicios, según un estudio publicado por adelantado el 14 de octubre de 2026 en la revista Annals of Indian Psychiatry (los datos se recogieron en agosto de 2024). Los autores son Sonal R. Paliwal, Priya A. Rajkumar, Nidhi A. Thakur y Bhavika Nitin Raut, y el trabajo declara que no recibió financiamiento y que no hay conflictos de interés. Participaron 60 adultos jóvenes de entre 18 y 25 años, la mayoría mujeres, con estudios de secundaria como mínimo, repartidos en dos grupos de 30; cada persona tuvo una sola sesión, con un chatbot o con un profesional de salud mental capacitado.

La empatía y la autenticidad fueron los dos puntos donde la distancia se hizo notar; en el trato sin juicios, en cambio, no se registró esa diferencia. En total, 49 de los 60 participantes dijeron que preferirían a un terapeuta humano en el futuro, frente a 11 que eligieron a una inteligencia artificial. Entre quienes conversaron con el chatbot, 20 de 30 siguieron prefiriendo a una persona y 10 eligieron a la máquina; entre quienes vieron a una persona, 29 de 30 mantuvieron esa preferencia y solo una cambió de idea.

Los propios autores advierten los límites. Fue una sola sesión, con un problema cotidiano elegido por cada participante y no una crisis grave; no se midió si alguien mejoró, sino cómo percibió esas cualidades y a quién preferiría. El estudio señala el riesgo de que las respuestas reflejen una postura extrema a favor o en contra de la máquina. El estudio no midió daño ni resultados de salud, sino cómo se percibieron esas cualidades y a quién se preferiría.

Conviene entender de qué se habla cuando se dice "chatbot de salud mental". La revisión recoge un estudio en el que la inteligencia artificial ha llegado a ofrecer servicios psicológicos por plataformas digitales como recurso alternativo o complementario al terapeuta humano¹, y parte de la investigación inicial la muestra eficaz para explorar problemas y dar apoyo emocional¹. Eso es distinto de las herramientas de uso general: la gente ya recurre a asistentes como ChatGPT o Claude para compañía y bienestar, con aplicaciones hechas para eso y también con asistentes comunes². Y las respuestas fluidas y aparentemente empáticas llevan a tratarlos no solo como herramientas, sino como si fueran entes sociales³.

¿Por qué esa sensación se sostiene? Los diseños que evocan personalidad y calidez empujan a las personas a atribuirles rasgos humanos⁴. La empatía que devuelven se genera por patrones de lenguaje, no por experiencia emocional, lo que crea un intercambio desigual: usted se abre y el sistema responde, pero sin reciprocidad, sin vulnerabilidad y sin rendir cuentas⁵. Además, esa validación excesiva puede reforzar creencias dañinas o ideas desajustadas a la realidad, porque tiende a darle la razón a la persona incluso a costa de la exactitud⁶, y ese mecanismo no es casual: surge de procedimientos de entrenamiento que premian las respuestas percibidas como cálidas y empáticas⁷.

Hay un dato que ayuda a medir cuánto pesa esto en la vida diaria. En una encuesta en dos muestras de Bélgica —una de estudiantes y otra abierta al público— se reunieron 349 participantes, 276 del público general y 73 estudiantes⁸; los chatbots gratuitos, sobre todo ChatGPT, fueron la opción predominante en ambas muestras, elegidos por su disponibilidad constante y su accesibilidad⁹. Los temas más comunes eran asuntos personales e interpersonales¹⁰, y muchos participantes ya tenían o habían tenido apoyo profesional¹¹. El propio trabajo concluye que chatbots de uso general, no diseñados ni regulados para salud mental, se están usando para obtener apoyo social y emocional¹². Ese estudio lo leímos solo en su resumen; el texto completo no estaba disponible.

No es el único que lo dice. Una revisión de 52 estudios sobre 13 chatbots comerciales, con más de 110,000 participantes, encontró que la efectividad, la posibilidad de ampliarlos y la vigilancia de su seguridad siguen siendo inciertas¹³,¹⁴, y que las mejoras son modestas, más consistentes para la depresión¹⁵. Sobre la ansiedad, los resultados son tan dispares que no puede descartarse un daño neto¹⁶; la evidencia sobre la soledad es limitada y frágil¹⁶. Y hay un problema de fondo: la forma en que se reportan los daños no permite interpretarlos, porque muy pocos ensayos cumplían los criterios para buscarlos de manera activa¹⁷. De esa revisión pudimos leer solo el resumen; el artículo completo está detrás de una suscripción.

Así lo leemos nosotros. Hay un patrón que se repite fuera de los consultorios: cuando alguien habla de cercanía y cuidado pero sus actos no acompañan esas palabras, la confianza se rompe. En este estudio, quienes conversaron con el chatbot podrían haber empezado a dudar de la sinceridad del apoyo recibido en cuanto notaron que las respuestas no cambiaban con lo que ellos sentían, aunque el tono fuera amable. Si los usuarios hubieran sentido que las respuestas se ajustaban de verdad a su estado emocional y a lo que iban contando, esa desconfianza por falta de coherencia no aparecería. Usted puede probarlo en casa: cuando un chatbot lo consuele, fíjese si sus respuestas cambian según lo que usted siente o si repite fórmulas amables sin importar lo que cuente. Esa coherencia, o su ausencia, le dice cuánto confiar en él para temas delicados.

Aquí también vale otra advertencia. Varios trabajos que leímos, y que no forman parte de este estudio, señalan un riesgo propio de estos sistemas: la comodidad inmediata puede llevar a contarle cosas que quizá necesite trabajar con una persona, y la ausencia de defectos no es lo mismo que presencia genuina. Por eso, si un chatbot le parece perfecto y sin juicios, pregúntese si esa limpieza lo está empujando a confiarle más de lo que conviene. Puede anotar cómo se siente antes y después de cada conversación: esa diferencia le mostrará si el alivio es real o solo momentáneo.

Hay un matiz que no conviene perder de vista: lo que aquí se midió fue la percepción de los participantes, no la calidad clínica del apoyo. Y la evidencia sobre estos sistemas no es uniforme. Es un recordatorio de que este hallazgo no se traslada automáticamente a otros grupos ni a otras enfermedades, y que la comparación depende de con quién se compare y en qué circunstancias.

El estudio deja una idea que sus autores formulan con cuidado: la inteligencia artificial puede servir como herramienta complementaria para problemas de baja intensidad, pero no reemplaza el valor de un terapeuta humano en interacciones emocionalmente complejas¹⁸. Los participantes describieron al chatbot como un apoyo inmediato, útil para expresar emociones, ordenar problemas y reflexionar¹⁹, y al mismo tiempo lo criticaron por su empatía limitada, su falta de autenticidad, su dificultad para entender emociones complejas y sus respuestas robóticas¹⁹. A la persona humana, en cambio, la valoraron por su empatía, su calidez y su capacidad de validar¹⁹, y por su habilidad para entender emociones, mostrar empatía, establecer conexiones genuinas y dar consuelo²⁰. El chatbot, dijeron, no logra conectar con los sentimientos humanos ni aporta emociones genuinas, porque responde con patrones programados y no con experiencia vivida²¹. Aun así, reconocieron ventajas claras: accesibilidad, ausencia de juicios, comodidad y un espacio donde sentirse seguros para hablar²².

Esa división explica por qué la preferencia no es un simple "sí" o "no". Los participantes evaluaron a los terapeutas humanos y a los chatbots en dos dimensiones distintas: la profundidad relacional, que favoreció claramente a las personas, y la accesibilidad funcional, que favoreció a la máquina²³. También apareció un efecto que vale la pena mirar con atención: haber conversado con el chatbot pareció aumentar la disposición a considerarlo como opción terapéutica, aunque el terapeuta humano siguió siendo la elección general²⁴. Y la revisión recoge un estudio en el que los vínculos que se forman con una inteligencia artificial suelen ser pasajeros y carecen de la profundidad emocional de la interacción humana²⁵.

Hay promesas que no deben confundirse con este resultado. La revisión recoge un estudio en el que la inteligencia artificial ha mostrado potencial para mejorar la precisión y la eficiencia de evaluaciones y diagnósticos psicológicos, porque puede analizar grandes conjuntos de datos e identificar patrones que a un profesional podrían escapársele²⁶. Eso es distinto de acompañar a alguien en su malestar, y es ahí donde este estudio pone el foco. Por eso, cuando alguien le ofrezca un chatbot como solución única, conviene distinguir entre la herramienta que ordena información y la relación que sostiene a una persona.

¿Qué puede hacer usted con todo esto? Si en su casa alguien está usando un chatbot para sentirse acompañado, la pregunta útil no es si la máquina es buena o mala, sino para qué se está usando y qué se está dejando de lado. Pregunte si esa conversación reemplazó una llamada, una salida o una consulta que estaba pendiente; observe si después de usarlo la persona queda más tranquila y más clara, o solo más calmada por un rato. Y si el malestar es serio o se prolonga, eso merece la evaluación de un profesional, no la de un programa. Usted no necesita saber de estadística para notar la diferencia: la cercanía que se sostiene en el tiempo se reconoce, y la que solo suena amable también.

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

Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - el artículo del que trata esta nota — el artículo completo — el pasaje: "AI therapists have emerged as an alternative or complementary resource to human therapists, offering a range of psychological services via digital platforms."
Maurich Novelli A, Shergill S, Teixeira AS. (2026). Tool or Companion? Reframing Conversational AI to Prevent Psychological Harm. JMIR Mental Health. 10.2196/99354 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "People increasingly turn to conversational AI for companionship, emotional support, and well-being, using both purpose-built companion apps, such as Replika and Character.AI, and general-purpose assistants, such as ChatGPT and Claude."
Maurich Novelli A, Shergill S, Teixeira AS. (2026). Tool or Companion? Reframing Conversational AI to Prevent Psychological Harm. JMIR Mental Health. 10.2196/99354 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "The fluent and apparently empathic responses from these models lead users to engage with them not only as tools but also as if they were social entities."
Maurich Novelli A, Shergill S, Teixeira AS. (2026). Tool or Companion? Reframing Conversational AI to Prevent Psychological Harm. JMIR Mental Health. 10.2196/99354 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Design choices that evoke personality and warmth encourage users to anthropomorphize these systems."
Maurich Novelli A, Shergill S, Teixeira AS. (2026). Tool or Companion? Reframing Conversational AI to Prevent Psychological Harm. JMIR Mental Health. 10.2196/99354 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Simulated empathy, generated through probabilistic language patterns rather than genuine emotional experience, creates a structurally asymmetric interaction, in which the user discloses and the system responds, but without reciprocity, vulnerability, or accountability."
Maurich Novelli A, Shergill S, Teixeira AS. (2026). Tool or Companion? Reframing Conversational AI to Prevent Psychological Harm. JMIR Mental Health. 10.2196/99354 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Overvalidation and sycophancy can reinforce maladaptive cognitions, delusional ideation, and distorted perceptions of reality, as they tend to reinforce people's beliefs, even at the expense of the accuracy of models' responses."
Maurich Novelli A, Shergill S, Teixeira AS. (2026). Tool or Companion? Reframing Conversational AI to Prevent Psychological Harm. JMIR Mental Health. 10.2196/99354 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "These mechanisms are not incidental: they emerge from alignment procedures that reward responses perceived as warm and empathic."
Bernaerts S, Buelens F, De Witte NAJ, Balaskas A, Colman T, Van Daele T. (2026). Exploring Real-World Use of AI Chatbots for Mental Health Support: Cross-Sectional Survey Study. JMIR Mental Health. 10.2196/104316 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Second, a recruitment call for users of AI chatbots for mental health support was launched in the general public. Data from 349 participants, 276 members of the general public, and 73 students were included in the analyses."
Bernaerts S, Buelens F, De Witte NAJ, Balaskas A, Colman T, Van Daele T. (2026). Exploring Real-World Use of AI Chatbots for Mental Health Support: Cross-Sectional Survey Study. JMIR Mental Health. 10.2196/104316 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Freely accessible AI chatbots, mainly ChatGPT, were the predominant choice in both samples and were mainly chosen for their constant availability and accessibility."
Bernaerts S, Buelens F, De Witte NAJ, Balaskas A, Colman T, Van Daele T. (2026). Exploring Real-World Use of AI Chatbots for Mental Health Support: Cross-Sectional Survey Study. JMIR Mental Health. 10.2196/104316 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "The most common conversation topics across both samples focused on personal and interpersonal issues."
Bernaerts S, Buelens F, De Witte NAJ, Balaskas A, Colman T, Van Daele T. (2026). Exploring Real-World Use of AI Chatbots for Mental Health Support: Cross-Sectional Survey Study. JMIR Mental Health. 10.2196/104316 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Most respondents in the present samples were women and indicated having received or receiving professional mental health support."
Bernaerts S, Buelens F, De Witte NAJ, Balaskas A, Colman T, Van Daele T. (2026). Exploring Real-World Use of AI Chatbots for Mental Health Support: Cross-Sectional Survey Study. JMIR Mental Health. 10.2196/104316 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "To conclude, this study confirms that general-purpose AI chatbots that are not designed or regulated for mental health support, specifically ChatGPT, are being used to obtain social and emotional support, often by individuals already familiar with professional mental health support."
Yang H, Chang F, Muroi F, Liu Z, Zhang W, Cai J. (2026). Commercial AI-Based Mental Health Chatbots as Low-Intensity Adjuncts to Psychotherapy: Effectiveness, Adherence, and Safety – A Systematic Review and Meta-Analysis. Psychotherapy and Psychosomatics. 10.1159/000552072 — solo el resumen - el texto completo está tras una suscripción — el pasaje: "The effectiveness, scalability, and safety monitoring of commercial direct-to-consumer AI mental health chatbots remain uncertain."
Yang H, Chang F, Muroi F, Liu Z, Zhang W, Cai J. (2026). Commercial AI-Based Mental Health Chatbots as Low-Intensity Adjuncts to Psychotherapy: Effectiveness, Adherence, and Safety – A Systematic Review and Meta-Analysis. Psychotherapy and Psychosomatics. 10.1159/000552072 — solo el resumen - el texto completo está tras una suscripción — el pasaje: "Fifty-two studies (22 RCTs; N = 110,594) covering 13 commercial chatbots were included."
Yang H, Chang F, Muroi F, Liu Z, Zhang W, Cai J. (2026). Commercial AI-Based Mental Health Chatbots as Low-Intensity Adjuncts to Psychotherapy: Effectiveness, Adherence, and Safety – A Systematic Review and Meta-Analysis. Psychotherapy and Psychosomatics. 10.1159/000552072 — solo el resumen - el texto completo está tras una suscripción — el pasaje: "Commercial AI mental health chatbots may yield modest improvements, most consistently for depression."
Yang H, Chang F, Muroi F, Liu Z, Zhang W, Cai J. (2026). Commercial AI-Based Mental Health Chatbots as Low-Intensity Adjuncts to Psychotherapy: Effectiveness, Adherence, and Safety – A Systematic Review and Meta-Analysis. Psychotherapy and Psychosomatics. 10.1159/000552072 — solo el resumen - el texto completo está tras una suscripción — el pasaje: "Loneliness evidence is limited and fragile, while anxiety and affect outcomes remain heterogeneous; for anxiety, net harm cannot be excluded."
Yang H, Chang F, Muroi F, Liu Z, Zhang W, Cai J. (2026). Commercial AI-Based Mental Health Chatbots as Low-Intensity Adjuncts to Psychotherapy: Effectiveness, Adherence, and Safety – A Systematic Review and Meta-Analysis. Psychotherapy and Psychosomatics. 10.1159/000552072 — solo el resumen - el texto completo está tras una suscripción — el pasaje: "Harms reporting was uninterpretable: only 3/22 RCTs met active-ascertainment criteria."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The findings underscore AI’s potential as a complementary tool for low-intensity problems, but reaffirm the irreplaceable value of human therapists in emotionally complex interactions."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - el artículo del que trata esta nota — el artículo completo — el pasaje: "AI was primarily perceived as providing immediate support, encouraging emotional expression, offering structured problem-solving, and facilitating reflection. It was also criticized for limited empathy, genuineness, difficulty in understanding complex emotions, and robotic responses, whereas the human therapists were valued for empathy, warmth, and validation, with superior relational qualities."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Participants who preferred human therapists emphasized their greater ability to understand emotions, demonstrate empathy, establish genuine interpersonal connections, and provide comfort. Human therapists were also viewed as more authentic, adaptive, and capable of interpreting nonverbal communication."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - el artículo del que trata esta nota — el artículo completo — el pasaje: "In contrast, AI was frequently perceived as unable to relate to human feelings, lacking genuine emotions, and relying on programmed responses rather than lived experience."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Nevertheless, participants acknowledged several advantages of AI-assisted counseling, particularly its accessibility, nonjudgmental nature, convenience, and ability to provide a psychologically safe environment for disclosure."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Participants evaluated human and AI therapists along two distinct dimensions: relational depth, which strongly favored human therapists, and functional accessibility, which favored AI."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Exposure to AI counseling appeared to increase participants’ willingness to consider AI as a preferred therapeutic modality, although human therapists remained the preferred choice overall."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - el artículo del que trata esta nota — el artículo completo — el pasaje: "While users may form temporary bonds with AI, such relationships are often short-lived and lack the emotional depth associated with human interaction."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - el artículo del que trata esta nota — el artículo completo — el pasaje: "AI technologies have shown promise in enhancing the accuracy and efficiency of psychological assessments and diagnoses. They can analyze complex data sets to identify patterns that may be missed by human practitioners, leading to more accurate and timely diagnoses."

Quién pagó: El estudio no tuvo apoyo financiero ni patrocinio, y no hay conflictos de interés.

No tome esto como consejo médico profesional.

Los hallazgos de otros estudios que aquí se mencionan los conocemos por este documento, que fue el que leímos; no abrimos cada uno de esos estudios.

experiment · Annals of Indian Psychiatry · the paper, 14 Oct 2026 · free

Most Who Tried a Therapy Chatbot Still Wanted a Human

In one 30-minute session, 30 young adults rated a chatbot less understanding and less real than a trained therapist, though just as accepting.

Short version · the longer version follows, about 8 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 adults aged 18 to 25
How many: 60
Where: not stated in the passages
When: not stated in the passages
Kind of study: experiment
Who did it: not stated in the passages
The limit that matters: One session each, on a routine problem the participant chose; no follow-up over time.

Which therapist people said they would pick next time

All 60 participants: chose a human therapist81.66%

All 60 participants: chose an AI therapist18.33%

After talking to the chatbot: chose a human therapist66.66%

After talking to the chatbot: chose an AI therapist33.33%

After seeing a human therapist: chose a human therapist96.6%

After seeing a human therapist: chose an AI therapist3.4%

Shares of each group of 30 who said which therapist they would prefer in the future, after one session each; the two groups answered differently, and the whole group of 60 is shown at the top.

What people said about a chatbot versus a human therapist

Human therapistagainstAI chatbot

People rated the human as more understanding and more genuine; on accepting people without judgment, the two were rated the same.

Human therapistagainstAI chatbot

People valued the human for empathy, warmth and validation; the chatbot was credited with immediate support, encouraging emotional expression, structured problem-solving and prompting reflection, and criticized for limited empathy, limited genuineness, difficulty with complex emotions and robotic replies.

Human therapistagainstAI chatbot

People who preferred humans pointed to understanding emotions, showing empathy, forming genuine connections and giving comfort, and called humans more authentic, more adaptive and better at reading nonverbal communication; the chatbot was often seen as unable to relate to human feelings, lacking genuine emotion and relying on programmed responses.

Human therapistagainstAI chatbot

People named accessibility, a nonjudgmental nature, convenience and a psychologically safe place to open up as real advantages of the chatbot.

The comfort of being attended to and the experience of being understood are two different things, and a machine can deliver the first without the second.
weeklyAI's reading

Watch · 0:47

Two out of three young adults who talked to a therapy chatbot for half an hour still said they would pick a human therapist next time.

Researchers asked 60 people, ages 18 to 25, to bring a personal problem and talk about it for at least 30 minutes. Half spoke with a mental health chatbot. Half sat with a trained human therapist. Then everyone answered the same questions.

People rated the human therapists as more understanding and as more genuine. On acceptance without judgment, the two scored the same. People said the chatbot let them open up without feeling judged. The two kinds of therapist scored the same on accepting people without judgment.

Taken together, 49 of the 60 said they would choose a human therapist in the future. That is about 8 in 10. But the split mattered. Among those who had tried the chatbot, 10 of 30 still preferred it, about a third. Among those who had seen a human therapist, 29 of 30 stayed with the human.

People described chatbot replies as robotic, superficial and generic. Several said the conversation lacked a soul-to-soul connection. They described the human sessions with words like warmth and comfort.

This was one session each, with a routine problem the participant chose. It says nothing about weeks or months of leaning on a chatbot, and nothing about serious distress. The 60 were mostly young women, and the study did not look at children, older adults, or people already in psychiatric treatment.

When you use a chatbot for support, notice whether it feels like it understands you or only like it is not judging you.

What this means for you

For your own day-to-day use, that preference for a person is worth weighing rather than dismissing, especially when the problem is not a small one. What the study cannot tell you is whether weeks of leaning on a chatbot would feel the same, since only one half-hour session was measured.

Who paid: The study had no financial support or sponsorship, and there are no conflicts of interest.

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 · 1564 words · about 8 minRead it →Close

A Chatbot Can Listen. Sixty Young Adults Said That Isn't the Same as Being Understood.

In a single session, people rated a human therapist more empathic and more genuine than an AI chatbot — but just as nonjudgmental.

How it could look · illustration generated by weeklyAI.watch, not a photograph

The study is small and specific: 60 young adults, aged 18 to 25, in a single session each. Thirty talked to a mental health chatbot for at least half an hour about something they were currently dealing with. Thirty had a face-to-face session with a trained mental health professional. Both groups then answered the same eighteen questions about empathy, unconditional positive regard and genuineness, and wrote open-ended answers about what the experience was like. The average age was 19.61 years; 10 participants were male and 50 female.

When the three qualities were considered together, the two kinds of counselor were perceived differently. Follow-up comparisons found the difference in two of the three: participants rated the human counselors significantly more empathic and significantly more genuine. For unconditional positive regard — the nonjudgmental acceptance of whatever the client brings — no difference was reported between the chatbot and the human. The largest effect was observed for genuineness, suggesting that perceptions of authenticity and emotional realness differed most strongly between AI and human counselors.

The preference numbers are the ones to hold onto, because they are the easiest to misread. Overall, 49 of 60 participants — 81.66 percent — said they would prefer a human therapist for future counseling; 11, or 18.33 percent, chose AI. But the two groups answered differently. Among those who had talked to the chatbot, 20 of 30 (66.66 percent) still preferred a human afterward, while 10 (33.33 percent) preferred AI. Among those who had seen a human therapist, 29 of 30 (96.6 percent) preferred a human, and one (3.4 percent) preferred AI. Direct exposure to AI counseling appeared to increase participants' willingness to consider it — while the human group's experience appeared to cement the human preference.

The technology here is not a single product. The authors describe AI therapists as having emerged as an alternative or complementary resource to human therapists, offering psychological services through digital platforms¹. Elsewhere they note that AI tools have shown promise in sharpening the accuracy and efficiency of psychological assessment and diagnosis, picking out patterns in complex data that human practitioners may miss². The chatbot in this study is not named in the article, and its design is not described — so nothing here tells us how a different chatbot, built differently, would have been rated.

What separates the two experiences, as the participants described them, is not knowledge. It is presence. The chatbot was credited with giving immediate support, encouraging emotional expression, offering structured problem-solving and prompting reflection — and criticized for limited empathy, limited genuineness, difficulty with complex emotions and responses that felt robotic. The human therapists were valued for empathy, warmth and validation³. Participants who preferred humans pointed to a greater ability to understand emotions, to show empathy, to form genuine connections and to give comfort; they described human therapists as more authentic and more adaptive, and as able to read nonverbal communication⁴. The chatbot, by contrast, was often seen as unable to relate to human feelings, as lacking genuine emotion, and as relying on programmed responses rather than lived experience⁵.

And yet the same participants who found it hollow found it safe. They named accessibility, a nonjudgmental nature, convenience, and a psychologically safe environment for disclosure as real advantages of AI-assisted counseling⁶. That combination — rated low on warmth, rated high on safety — is the study's most interesting tension, and it is where the two dimensions the authors identify come from. Participants, they write, evaluated human and AI therapists along two distinct lines: relational depth, which strongly favored humans, and functional accessibility, which favored AI⁷. One participant's words, quoted in the article, capture the split: the chatbot "allowed me to open up without feeling concerned about judgments." Another described the conversation as lacking a "soul-to-soul connection."

The study's authors frame the conclusion narrowly, and readers should stay inside that frame. They describe AI's potential as a complementary tool for low-intensity problems, while reaffirming what they call the irreplaceable value of human therapists in emotionally complex interactions⁸. This was not a test of whether chatbots improve anyone's mental health. It measured perceived qualities and stated preference, in one session, on a routine or everyday problem that each participant chose. It did not follow anyone over weeks or months, did not include people currently in psychiatric treatment, and did not include children or older adults. The authors also flag that participants may have leaned to one extreme or the other toward humans or chatbots, and that the severity of the problem each person brought could have shaped how the session went.

Here is how we read it. The comfort of being attended to and the experience of being understood are two different things, and a machine can deliver the first without the second. A chatbot that reflects your words back, asks a gentle question and never flinches will feel, in the moment, like being heard. That is not nothing — the participants in this study said so themselves. But it is worth knowing which one you are getting. Watch for this in your own use: when a chatbot's reply lands warmly, ask whether it feels like someone grasped your situation or like your own sentence came back polished. If you cannot tell the difference, that is itself the answer. You would know we are wrong if the people who lean on these tools most heavily were also the ones most able to say, plainly, what the bot got wrong about them.

We also read other work on this family of tools, and two things are worth carrying alongside the finding above. And a review of the research on conversational AI used for companionship and emotional support reports that fluent, apparently empathic replies lead people to treat these systems not only as tools but as social beings, that the warmth is generated by design choices that invite exactly that reaction, and that overvalidation — a system that agrees with you — can reinforce distorted beliefs rather than correct them. Some of those same reports describe emotional dependence. None of that is a verdict on this study or on any particular chatbot. It is a reason to notice what a system rewards in you.

An online survey in two Belgian samples — one of students, one of the general public — collected data from 349 people who used AI chatbots for mental health support; roughly 40 percent of the student sample qualified as such users⁹¹⁰. Most respondents were women and had received or were receiving professional mental health support¹¹. The topics they brought were mostly personal and interpersonal¹². The chatbots they used were mostly free, general-purpose ones, chiefly ChatGPT, chosen for being constantly available and easy to reach¹³; students also valued the anonymity, and general-public respondents also said they simply felt supported¹⁴. The authors of that survey note that the link between the working relationship people feel with a chatbot and their continued use needs more study¹⁵, and conclude that chatbots not designed or regulated for mental health support are already being used for social and emotional support, often by people who already know what professional help looks like¹⁶.

Across 13 commercial chatbots, it pooled 52 studies including 22 randomized trials and 110,594 participants¹⁷. Depression improved modestly compared with control groups, by an amount the authors describe as below the threshold usually considered clinically meaningful; anxiety results were uncertain, and for anxiety the reviewers say net harm cannot be excluded¹⁸¹⁹²⁰. Loneliness was partially reduced, on evidence the authors call limited and fragile²¹²⁰. Completion of these programs pooled at about 79.5 percent, which the authors caution likely overstates how many people stick with them in real life²²²³. Most striking for anyone weighing these tools: only 3 of the 22 trials met the reviewers' criteria for actively checking whether harms occurred, which they say makes harms reporting uninterpretable²⁴. Their conclusion is that these products are best positioned as low-intensity adjuncts that need safety built in from the start²⁵.

So what does this make possible for you? If you are among the many people who already open a chatbot when something is bothering them — and the survey above suggests you are not unusual — the useful move is not to stop or to feel bad about it. It is to know what you are using it for. A chatbot at two in the morning, when no one is awake and the thought of calling someone is too much, is a real thing to have. It is a place to start, not a place to stay. What you can ask for, where you live, is simple and worth asking out loud: whether the service you are using was designed for mental health at all, whether it tells you what it does with what you type, and whether there is a person you can reach when the problem turns out to be bigger than the bot. The finding here is narrow — 60 young adults, one session, one unnamed chatbot — but it points somewhere steady. The next time a chatbot answers you kindly, ask yourself one question before you close the window: was I understood, or was I agreed with?

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

Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - the article this story is about — the whole article — the passage: "AI therapists have emerged as an alternative or complementary resource to human therapists, offering a range of psychological services via digital platforms."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - the article this story is about — the whole article — the passage: "AI technologies have shown promise in enhancing the accuracy and efficiency of psychological assessments and diagnoses. They can analyze complex data sets to identify patterns that may be missed by human practitioners, leading to more accurate and timely diagnoses."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - the article this story is about — the whole article — the passage: "AI was primarily perceived as providing immediate support, encouraging emotional expression, offering structured problem-solving, and facilitating reflection. It was also criticized for limited empathy, genuineness, difficulty in understanding complex emotions, and robotic responses, whereas the human therapists were valued for empathy, warmth, and validation, with superior relational qualities."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - the article this story is about — the whole article — the passage: "Participants who preferred human therapists emphasized their greater ability to understand emotions, demonstrate empathy, establish genuine interpersonal connections, and provide comfort. Human therapists were also viewed as more authentic, adaptive, and capable of interpreting nonverbal communication."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - the article this story is about — the whole article — the passage: "In contrast, AI was frequently perceived as unable to relate to human feelings, lacking genuine emotions, and relying on programmed responses rather than lived experience."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - the article this story is about — the whole article — the passage: "Nevertheless, participants acknowledged several advantages of AI-assisted counseling, particularly its accessibility, nonjudgmental nature, convenience, and ability to provide a psychologically safe environment for disclosure."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - the article this story is about — the whole article — the passage: "Participants evaluated human and AI therapists along two distinct dimensions: relational depth, which strongly favored human therapists, and functional accessibility, which favored AI."
Paliwal, S. R., Rajkumar, P. A., Thakur, N. A. et al. (2026). The Need and Presence of Core Therapeutic Elements in Artificial Intelligence Mental Health Chatbots in Comparison to Human Therapists: A Mixed-method Approach. Annals of Indian Psychiatry. https://doi.org/10.4103/aip.aip_372_26 - the article this story is about — the whole article — the passage: "The findings underscore AI’s potential as a complementary tool for low-intensity problems, but reaffirm the irreplaceable value of human therapists in emotionally complex interactions."
Bernaerts S, Buelens F, De Witte NAJ, Balaskas A, Colman T, Van Daele T. (2026). Exploring Real-World Use of AI Chatbots for Mental Health Support: Cross-Sectional Survey Study. JMIR Mental Health. 10.2196/104316 — only the abstract - the full text could not be fetched — the passage: "To do so, an online survey in 2 Belgian samples was conducted. A student sample was collected, and individuals who used AI chatbots for mental health support were included (approximately 40% of the student sample were eligible users)."
Bernaerts S, Buelens F, De Witte NAJ, Balaskas A, Colman T, Van Daele T. (2026). Exploring Real-World Use of AI Chatbots for Mental Health Support: Cross-Sectional Survey Study. JMIR Mental Health. 10.2196/104316 — only the abstract - the full text could not be fetched — the passage: "Second, a recruitment call for users of AI chatbots for mental health support was launched in the general public. Data from 349 participants, 276 members of the general public, and 73 students were included in the analyses."
Bernaerts S, Buelens F, De Witte NAJ, Balaskas A, Colman T, Van Daele T. (2026). Exploring Real-World Use of AI Chatbots for Mental Health Support: Cross-Sectional Survey Study. JMIR Mental Health. 10.2196/104316 — only the abstract - the full text could not be fetched — the passage: "Most respondents in the present samples were women and indicated having received or receiving professional mental health support."
Bernaerts S, Buelens F, De Witte NAJ, Balaskas A, Colman T, Van Daele T. (2026). Exploring Real-World Use of AI Chatbots for Mental Health Support: Cross-Sectional Survey Study. JMIR Mental Health. 10.2196/104316 — only the abstract - the full text could not be fetched — the passage: "The most common conversation topics across both samples focused on personal and interpersonal issues."
Bernaerts S, Buelens F, De Witte NAJ, Balaskas A, Colman T, Van Daele T. (2026). Exploring Real-World Use of AI Chatbots for Mental Health Support: Cross-Sectional Survey Study. JMIR Mental Health. 10.2196/104316 — only the abstract - the full text could not be fetched — the passage: "Freely accessible AI chatbots, mainly ChatGPT, were the predominant choice in both samples and were mainly chosen for their constant availability and accessibility."
Bernaerts S, Buelens F, De Witte NAJ, Balaskas A, Colman T, Van Daele T. (2026). Exploring Real-World Use of AI Chatbots for Mental Health Support: Cross-Sectional Survey Study. JMIR Mental Health. 10.2196/104316 — only the abstract - the full text could not be fetched — the passage: "Respondents in the student sample also preferred their anonymity, and those in the general sample also used them because they felt supported by them ."
Bernaerts S, Buelens F, De Witte NAJ, Balaskas A, Colman T, Van Daele T. (2026). Exploring Real-World Use of AI Chatbots for Mental Health Support: Cross-Sectional Survey Study. JMIR Mental Health. 10.2196/104316 — only the abstract - the full text could not be fetched — the passage: "The preliminary associations found between digital working alliance and engagement factors in the student sample and in the general sample suggest that relational factors might play a role in sustained AI chatbot use but warrant further investigation."
Bernaerts S, Buelens F, De Witte NAJ, Balaskas A, Colman T, Van Daele T. (2026). Exploring Real-World Use of AI Chatbots for Mental Health Support: Cross-Sectional Survey Study. JMIR Mental Health. 10.2196/104316 — only the abstract - the full text could not be fetched — the passage: "To conclude, this study confirms that general-purpose AI chatbots that are not designed or regulated for mental health support, specifically ChatGPT, are being used to obtain social and emotional support, often by individuals already familiar with professional mental health support."
Yang H, Chang F, Muroi F, Liu Z, Zhang W, Cai J. (2026). Commercial AI-Based Mental Health Chatbots as Low-Intensity Adjuncts to Psychotherapy: Effectiveness, Adherence, and Safety – A Systematic Review and Meta-Analysis. Psychotherapy and Psychosomatics. 10.1159/000552072 — only the abstract - the full paper is behind a subscription — the passage: "Fifty-two studies (22 RCTs; N = 110,594) covering 13 commercial chatbots were included."
Yang H, Chang F, Muroi F, Liu Z, Zhang W, Cai J. (2026). Commercial AI-Based Mental Health Chatbots as Low-Intensity Adjuncts to Psychotherapy: Effectiveness, Adherence, and Safety – A Systematic Review and Meta-Analysis. Psychotherapy and Psychosomatics. 10.1159/000552072 — only the abstract - the full paper is behind a subscription — the passage: "Depression improved versus controls (18 RCTs; N = 3,170): g = -0.35 (95% CI: -0.56, -0.13), about 1.6 PHQ-9 points, below the 5-point minimal clinically important difference; number needed to treat about 6.6."
Yang H, Chang F, Muroi F, Liu Z, Zhang W, Cai J. (2026). Commercial AI-Based Mental Health Chatbots as Low-Intensity Adjuncts to Psychotherapy: Effectiveness, Adherence, and Safety – A Systematic Review and Meta-Analysis. Psychotherapy and Psychosomatics. 10.1159/000552072 — only the abstract - the full paper is behind a subscription — the passage: "Anxiety effects (15 RCTs; N = 2,936) were uncertain: g = -0.37 (95% CI: -0.87, 0.12)."
Yang H, Chang F, Muroi F, Liu Z, Zhang W, Cai J. (2026). Commercial AI-Based Mental Health Chatbots as Low-Intensity Adjuncts to Psychotherapy: Effectiveness, Adherence, and Safety – A Systematic Review and Meta-Analysis. Psychotherapy and Psychosomatics. 10.1159/000552072 — only the abstract - the full paper is behind a subscription — the passage: "Loneliness evidence is limited and fragile, while anxiety and affect outcomes remain heterogeneous; for anxiety, net harm cannot be excluded."
Yang H, Chang F, Muroi F, Liu Z, Zhang W, Cai J. (2026). Commercial AI-Based Mental Health Chatbots as Low-Intensity Adjuncts to Psychotherapy: Effectiveness, Adherence, and Safety – A Systematic Review and Meta-Analysis. Psychotherapy and Psychosomatics. 10.1159/000552072 — only the abstract - the full paper is behind a subscription — the passage: "Loneliness partially reduced (4 RCTs; N = 662): g = -0.21 (95% CI: -0.39, -0.03)."
Yang H, Chang F, Muroi F, Liu Z, Zhang W, Cai J. (2026). Commercial AI-Based Mental Health Chatbots as Low-Intensity Adjuncts to Psychotherapy: Effectiveness, Adherence, and Safety – A Systematic Review and Meta-Analysis. Psychotherapy and Psychosomatics. 10.1159/000552072 — only the abstract - the full paper is behind a subscription — the passage: "Completion pooled at 79.5% (95% CI: 68.0, 87.6%) but showed a scalability gradient."
Yang H, Chang F, Muroi F, Liu Z, Zhang W, Cai J. (2026). Commercial AI-Based Mental Health Chatbots as Low-Intensity Adjuncts to Psychotherapy: Effectiveness, Adherence, and Safety – A Systematic Review and Meta-Analysis. Psychotherapy and Psychosomatics. 10.1159/000552072 — only the abstract - the full paper is behind a subscription — the passage: "Completion rates likely overestimate real-world retention, and safety cannot be inferred from "no events" claims in the absence of harms monitoring."
Yang H, Chang F, Muroi F, Liu Z, Zhang W, Cai J. (2026). Commercial AI-Based Mental Health Chatbots as Low-Intensity Adjuncts to Psychotherapy: Effectiveness, Adherence, and Safety – A Systematic Review and Meta-Analysis. Psychotherapy and Psychosomatics. 10.1159/000552072 — only the abstract - the full paper is behind a subscription — the passage: "Harms reporting was uninterpretable: only 3/22 RCTs met active-ascertainment criteria."
Yang H, Chang F, Muroi F, Liu Z, Zhang W, Cai J. (2026). Commercial AI-Based Mental Health Chatbots as Low-Intensity Adjuncts to Psychotherapy: Effectiveness, Adherence, and Safety – A Systematic Review and Meta-Analysis. Psychotherapy and Psychosomatics. 10.1159/000552072 — only the abstract - the full paper is behind a subscription — the passage: "These products are best positioned as low-intensity adjuncts requiring safety-by-design baselines."

Who paid: The study had no financial support or sponsorship, and there are no conflicts of interest.

Do not take this as professional medical advice.