Most Future Teachers Reach for AI First. A Third Can't Be Bothered to Check What It Says.

survey · Frontiers in psychology · la publicación, 12 ago 2026 · gratis

Un estudio a futuros maestros en China halló que la impulsividad y el apego emocional a la IA predicen mayor dependencia

Una encuesta a estudiantes de magisterio en China encuentra dependencia extendida; los rasgos que más la acompañan no son la confianza en la herramienta, sino la impulsividad y el apego afectivo hacia ella.

Versión breve · la versión detallada sigue, unos 6 min

El estudio, de un vistazo
Quiénes: Estudiantes de carreras de educación en programas de formación docente
Cuántos: 565 estudiantes
Dónde: China
Cuándo: No lo dice el estudio
Tipo de estudio: survey
Quién lo hizo: Universidad de Qingdao
El límite que importa: Un solo momento en el tiempo y todo medido con autoreportes: son asociaciones, no causas demostradas.

Cuántos futuros maestros muestran cada señal de dependencia de la IA

Dicen que la IA se volvió indispensable para estudiar79%

Se sienten desacostumbrados tras un tiempo sin usarla78%

Su primera reacción ante una tarea es preguntarle a la IA54%

Usan sus respuestas sin verificar nada31%

Temen que depender a largo plazo deteriore su capacidad de enseñar44%

Querrían reducir su dependencia47%

Porcentajes de 565 estudiantes chinos de magisterio que estuvieron de acuerdo con cada afirmación; describen a ese grupo en un solo momento, no a todos los futuros docentes.

Cómo hablaban de la IA los entrevistados más y menos dependientes

El entrevistado que más personificaba a la IAfrente aLos dos que insistían en que la IA es solo una herramienta

El primero tuvo más dificultad para trabajar sin ella; los otros dos fueron los menos dependientes.

Estudiantes de primero, más confiados en que la IA rara vez se equivocafrente aEstudiantes de años mayores, que detectaban citas inventadas

Los mayores consideraban utilizable apenas entre 20% y 30% de lo que la herramienta produce.

La próxima vez que usted delegue algo en una máquina, ¿podría intentarlo primero sin ella?
Lectura de weeklyAI

Ver · 0:50

Ellos todavía no tienen alumnos. Estudian para ser maestros. Y en una encuesta, 79 de cada 100 dijeron que la inteligencia artificial ya es una parte indispensable de su aprendizaje. Setenta y ocho de cada 100 afirmaron sentirse desacostumbrados tras un tiempo sin usarla.

El estudio reunió a 565 estudiantes de carreras de educación en programas chinos de formación docente. Respondieron un cuestionario en línea y cinco de ellos fueron entrevistados después.

Los investigadores analizaron las respuestas con un modelo estadístico que busca el aporte propio de cada factor, con un análisis de condiciones necesarias y con una exploración de combinaciones de factores. El estudio lo realizaron Bin Jiang y Zhuo Wang, de la Universidad de Qingdao.

La dependencia apareció extendida. El 54 por ciento dijo que su primera reacción ante una tarea es preguntarle a la IA. El 44 por ciento admite entregarle a la IA incluso tareas que podría hacer solo. El 31 por ciento no se molesta en verificar sus conclusiones. El 34 por ciento adopta sus respuestas sin pensarlas mucho. Al mismo tiempo, 47 de cada 100 desearían reducir su dependencia.

Los factores que más acompañaron esa dependencia fueron la impulsividad —preferir el resultado rápido al esfuerzo— y el apego afectivo: tratar a la IA como un compañero o un amigo. La confianza en la exactitud de la herramienta no mostró una relación clara.

El diseño es transversal y todo se midió con autoreportes. Los resultados describen asociaciones en un momento dado, no causas demostradas. Nadie puede afirmar todavía que enseñar autorregulación vaya a reducir la dependencia.

Sobre la protección, el hallazgo es matizado. La alfabetización en IA percibida por los propios estudiantes ayudó, pero poco. Su aporte protector solo se vio al separarla de otros factores, porque quienes se sentían más competentes también confiaban más en la IA y convivían con compañeros más dependientes de ella.

Dos advertencias sobre esa medida. La alfabetización fue autoevaluada, no puesta a prueba, y casi todos se calificaron alto. Un estudiante seguro de sus habilidades puede, aun así, delegar mucho. Además, la muestra se concentró en una sola región de China, con mayoría de mujeres y de estudiantes de primer año. Los números describen a ese grupo, no a sus estudiantes ni a sus hijos.

Los autores recomiendan que la formación docente enseñe autorregulación y que la alfabetización en IA se enseñe como complemento, no como única protección. A los futuros maestros les sugieren fijar límites sobre cuándo consultar la IA, intentar las tareas sin ella antes de recurrir a ella y verificar sus respuestas. También sugieren actividades breves para que los estudiantes noten cuándo tratan a un programa como si fuera un acompañante y no una herramienta.

Nada de esto está probado como intervención; son hipótesis por ensayar.

La próxima vez que usted delegue algo en una máquina, ¿podría intentarlo primero sin ella?

Qué significa para usted

Para usted, que aprende o enseña, la señal a observar no es cuánta IA usa, sino si prefiere el resultado rápido al esfuerzo y si conversa con el programa como con un compañero. No concluya todavía que enseñar autorregulación lo resuelva: el estudio solo muestra asociaciones en un momento dado, y sus entrevistados aún no tienen alumnos.

Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716

Quién pagó: La investigación fue financiada por el Departamento de Educación del Gobierno de China y por un proyecto de investigación de humanidades y ciencias sociales del Ministerio de Educación (subvención n.º 23JDSZ3046); el artículo no indica que los financiadores intervinieran en el diseño, el análisis o la redacción, ni que se prestara equipo o software.

Versión detalladaLos pasajes copiados del artículo, las ilustraciones y cada fuente con cuánto leímos de ella · 1267 palabras · unos 6 minLeerla →Cerrar

El 31% de los futuros profesores admite que no se molesta en verificar las respuestas de la IA

Un estudio con 565 estudiantes chinos de magisterio mide cuánto dependen de la inteligencia artificial generativa y qué los empuja a hacerlo.

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

La inteligencia artificial generativa —herramientas como ChatGPT, DeepSeek o Doubao, que redactan texto y responden preguntas en lenguaje corriente— se volvió algo cotidiano en la universidad. Un estudio de Bin Jiang y Zhuo Wang, de la Universidad de Qingdao, en China, preguntó qué tan metida está esa costumbre en quienes se están formando para ser maestros.¹ La respuesta es que ya no es un apoyo ocasional: 79% de los encuestados dijo que la IA se volvió indispensable para estudiar, 78% se siente desacostumbrado cuando pasa un tiempo sin ella, 54% reconoció que su primera reacción ante una tarea es preguntarle a la IA, y 31% usa sus respuestas sin verificar nada.²

Los investigadores no se quedaron en contar cuántos. Buscaron qué empuja esa dependencia. Lo que más pesó fue la impulsividad —la preferencia por lo rápido sobre lo esforzado— y el apego a la IA como si fuera alguien cercano. La ansiedad por quedarse atrás aportó menos, y la confianza en la precisión técnica de la herramienta no alcanzó significancia estadística.³ La alfabetización en IA que cada estudiante se atribuye apareció como protectora, pero débil, y su efecto se veía menor cuando se la miraba sola, porque quienes se sienten más competentes también confían más en la IA y se mueven en grupos donde todos la usan.⁴

Vale la pena entender de qué se trata esta tecnología antes de sacar conclusiones. La IA generativa —herramientas como ChatGPT, DeepSeek o Doubao— transformó cómo los estudiantes universitarios buscan información, redactan trabajos y preparan clases.¹ Esa respuesta inmediata es justamente lo que la vuelve tan fácil de incorporar al estudio diario.

El interés del estudio está en quiénes respondieron. No son estudiantes cualesquiera: son futuros docentes, y por eso el riesgo se duplica. Como estudiantes de hoy, están expuestos a la misma delegación de esfuerzo que sus compañeros; como educadores de mañana, los hábitos que formen ahora moldearán cómo la próxima generación de alumnos aprende a usar la IA o a depender de ella.⁵ Un maestro que no puede trabajar sin la herramienta difícilmente enseñe a pensar sin ella.

No es el primer trabajo sobre el tema. Ya existía una primera generación de estudios que empezaba a mapear qué lleva a depender de la IA, y en el caso de los docentes un trabajo previo encontró que lo que más empujaba la adicción era la absorción cognitiva, esa sensación de quedar inmerso en la herramienta.⁶ Este estudio nuevo apunta en la misma dirección: lo que engancha no es una evaluación fría de utilidad, sino procesos más automáticos y afectivos.

La parte cualitativa del estudio ayuda a ver el mecanismo. Entre cinco entrevistados, quien más personificaba a la IA —"quiero que sea mi compañera, mi amiga, mi maestra"— fue también quien más dificultad tuvo para trabajar sin ella. Los dos que insistían en que la IA es "solo una herramienta, las manos y los pies, mientras la persona es el cerebro" resultaron los menos dependientes.⁷ También apareció un contraste entre años de estudio: los de primero, más confiados en que la IA "rara vez se equivoca", frente a los mayores, que detectaban citas inventadas y consideraban utilizable apenas entre 20% y 30% de lo que la herramienta produce.⁸

Y no todo es comodidad. Un grupo importante de estos futuros maestros ya siente incomodidad profesional: 44% teme que depender a largo plazo deteriore su capacidad de enseñar, y 47% querría reducir su dependencia.⁹ Es decir, buena parte sabe que algo no está bien, aunque siga usando la herramienta igual.

Conviene ser claro sobre lo que este estudio no puede decir. Fue un solo momento en el tiempo, con encuestas y sin seguimiento: las relaciones que reporta son asociaciones, no causas demostradas, y algunas podrían funcionar en los dos sentidos —depender más podría a la vez profundizar el apego—.¹⁰ Tampoco es una muestra representativa de todos los estudiantes: 565 personas, 96.1% concentradas en la provincia de Shandong, 76.8% mujeres y 60.5% de primer año.¹¹ Además, la "alfabetización en IA" se midió por lo que cada uno dice de sí mismo, no con pruebas, y varios llegaban al tope de la escala: alguien puede sentirse muy capacitado y aun así delegar sin pensar.¹²

Así lo leemos nosotros. El hallazgo que más debería inquietar a quien enseña o cría hijos no es el número de usuarios, sino la combinación de dos cosas: la comodidad extrema y el afecto. Cuando una herramienta responde al instante, siempre está disponible y no juzga, es fácil empezar a tratarla como compañía y no como calculadora. Y esa cercanía afectiva, no la simple frecuencia de uso, es la que parece sostener la dependencia cuando la tarea se pone difícil.

Lo que esto nos lleva a esperar en una casa o un aula como la suya es concreto: los estudiantes que hablan de la IA como "alguien" —que le piden opinión sobre sí mismos, que sienten que los acompaña— serán los que más les cueste trabajar sin ella, más incluso que los que solo confían en su precisión técnica. Nos equivocaríamos si esos estudiantes que la describen como una herramienta fría mostraran igual o mayor dependencia, o si la cercanía afectiva no predijera nada una vez descontado el simple uso frecuente. Para saberlo usted no necesita un estudio: fíjese en el lenguaje. Escuche si la IA aparece como "compañera", "amiga", "alguien", y pregunte —a sus estudiantes, a sus hijos, a usted mismo— cuándo deciden no usarla y por qué. Esa pregunta revela más que cualquier encuesta de confianza.

Si estos resultados se sostienen, lo que sigue no es prohibir la herramienta. Los autores señalan que nada de esto justifica restringir el acceso, porque eso dejaría de lado los beneficios de un uso bien manejado.¹³ Lo que proponen es otro camino: enseñar autorregulación —fijarse metas, monitorear el propio avance, tolerar la dificultad de pensar sin ayuda— para que hacer una pausa antes de delegar se vuelva un hábito entrenado.¹³ Y como el apego también empuja, sugieren actividades breves que hagan visible el tirón emocional de los agentes que parecen humanos, por ejemplo notar cuándo se trata a un chatbot como compañero en vez de herramienta.¹⁴ La alfabetización en IA sigue siendo valiosa, pero como complemento, no como salvaguarda única: enseñarla sin apoyo de autorregulación podría aumentar el uso confiado sin frenar la dependencia.¹⁵

Hay una segunda cosa que este estudio deja a la vista y que vale la pena tener presente. La confianza que alguien dice tener en su propia destreza con la IA no es lo mismo que la destreza. Los datos sugieren que sentirse competente puede acompañarse de delegar más, no menos. Si eso fuera cierto, evaluar a un estudiante o a un colega por lo que sabe de IA diría poco; habría que mirar lo que hace cuando la herramienta no está disponible, porque ahí se revela la dependencia mejor que en una encuesta de confianza. Usted puede empezar por algo pequeño: proponer en su clase o en su casa una conversación regular sobre cuándo y cómo se usa la IA, y observar qué pasa un día sin ella. Eso no cuesta nada y muestra más que cualquier diagnóstico.

Lo que este estudio deja en sus manos es una pregunta sencilla para la próxima vez que alguien —usted, un estudiante, un hijo— abra la herramienta antes de pensar: ¿esto lo hago yo primero, o ya decidí que no puedo?

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

Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The rapid diffusion of generative artificial intelligence (GenAI) tools—such as ChatGPT, DeepSeek, and Doubao—has transformed how university students search for information, draft assignments, and prepare lessons."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Dependency markers were already widespread: 79% called AI indispensable to their learning, 78% felt unaccustomed without it, 54% “asked AI first,” and 31% used its answers without verification."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The strongest drivers were impulsivity (β = 0.40, p < 0.001) and anthropomorphic attachment (β = 0.28, p < 0.001), with a smaller significant contribution from AI anxiety (β = 0.15, p = 0.005), supporting H1, H2, and H4."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Consistent with H5, perceived AI literacy had a significant protective effect (β = −0.19, p < 0.001). This protective role was partly muted at the bivariate level (zero-order r = −0.14) because perceived literacy correlated positively with trust (r = 0.19) and social norms (r = 0.38), both associated with greater dependency."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - el artículo del que trata esta nota — el artículo completo — el pasaje: "For pre-service teachers, this risk is doubled. As current learners, they are exposed to the same offloading dynamics as their peers; as future educators, the habits and judgment they form now will shape how a next generation of pupils learns to use—or over-rely on—AI."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A first generation of empirical studies has begun to map the antecedents of AI dependency (Zhang et al., 2024; Li and Xing, 2025), and in the teacher context. Du et al. (2026) applied the I-PACE model and found cognitive absorption to be the strongest predictor of teachers’ GenAI addiction."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The clearest convergence concerned anthropomorphic attachment: the informant who most vividly personified AI (“I want AI to be my classmate, my friend, my teacher”) also reported the most difficulty working without it, whereas the two who insisted AI is “just a tool… the hands and feet, while the person is the brain” were the least reliant."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A trust-versus-verification gradient distinguished the more reliant first-years, who judged AI “rarely wrong,” from the critical upper-years, who caught fabricated citations and deemed only 20–30% of AI output usable."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - el artículo del que trata esta nota — el artículo completo — el pasaje: "At the same time, a substantial minority voiced professional unease: 44% worried that long-term reliance would erode their teaching ability, and 47% wished to reduce their dependence."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The design was cross-sectional and cannot establish temporal ordering or rule out reverse causality; the SEM paths are theoretically motivated associations, not demonstrated effects, and some relations are plausibly reciprocal (for example, reliance may both follow from and deepen anthropomorphic attachment)."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Because recruitment relied on an anonymous convenience and peer-referral strategy, we could not control the sample’s composition: it was geographically concentrated in Shandong (96.1%, chiefly Jinan and Qingdao), predominantly female (76.8%), and skewed toward first-year students (60.5%)."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - el artículo del que trata esta nota — el artículo completo — el pasaje: "All measures were self-reported, including perceived AI literacy, which by construction captures self-assessed rather than tested competence; future work should pair self-report with objective literacy tasks and behavioral indicators of reliance."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Because impulsivity was the strongest driver, the highest-yield strategy is likely to embed metacognitive self-regulation—goal setting, progress monitoring, and the deliberate tolerance of “desirable difficulties”—into coursework (Zimmerman, 1990; Bjork et al., 2013), so that pausing to think before delegating to AI becomes a trained habit."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Because anthropomorphic attachment also drove reliance, programs might add brief reflective activities that make the parasocial pull of human-like agents visible—for instance, prompting students to notice when they treat a chatbot as a companion rather than a tool (Skjuve et al., 2021)."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Perceived AI literacy should still be taught, but as a complement rather than a stand-alone safeguard: because more self-assured students also trust AI more and are more embedded in AI-using peer norms, literacy training delivered without self-regulation support risks raising confident use without curbing dependency."

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.

survey · Frontiers in psychology · the paper, 12 Aug 2026 · free

Most Future Teachers Reach for AI First. A Third Can't Be Bothered to Check What It Says.

In a survey of 565 Chinese education majors, dependence on AI was already the norm — and the students who rated their own AI skills highest were not the most protected.

Short version · the longer version follows, about 6 min

The study at a glance
Who: students training to become teachers
How many: 565
Where: China
When: 2026
Kind of study: survey
Who did it: Qingdao University
The limit that matters: One round of self-reported answers; cannot show cause

How common AI reliance was among future teachers

AI indispensable to learning79%

Felt unaccustomed without AI78%

Asked AI first54%

Used answers without verification31%

Share of 565 Chinese education students reporting each habit; all answers were self-reported in one survey.

Watch · 0:52

In teacher-training programs in China, 565 students preparing to stand in front of classrooms answered questions about how they study. Nearly eight in ten said AI had become an indispensable part of their learning. About the same share said they felt unaccustomed after a stretch without it. More than half said their first reaction to a task was to ask AI first.

The study — published in Frontiers in Psychology by Bin Jiang and Zhuo Wang of Qingdao University — used an online survey plus five follow-up interviews. The survey answers were analyzed three ways: to find what moved together with heavy AI use, to test whether any single factor had to be present for it to occur, and as an exploratory check on whether factors combined into distinct patterns.

Dependence markers were common. About a third of respondents said they adopted AI answers without thinking deeply, and 31 percent said they could not be bothered to verify AI's conclusions. Yet 71 percent believed AI's answers were generally accurate, and nearly half wished to reduce their own dependence.

What drove heavy reliance was not trust in AI's accuracy, which showed no clear relationship once other factors were accounted for. The strongest links were to impulsivity — choosing the fast route over the effortful one — and to anthropomorphic attachment: treating the chatbot as something like a companion. Anxiety about falling behind played a smaller role.

The design cannot show cause. This was one round of self-reported answers, with no experiment and no tracking of actual behavior, so these are associations. No one can yet say that teaching self-regulation would reduce dependence.

Perceived AI literacy — students' confidence in their own AI skills — was linked to somewhat lower dependence. That protective link was modest, and it was hidden at first glance: students who rated their skills highly also tended to trust AI more and to sit among peers who relied on it more, both of which went with heavier use.

That measure was self-rated, not tested, and scores crowded near the top, so a confident student may still hand work over wholesale. The sample was also concentrated in one Chinese province, mostly female and mostly first-year. The numbers describe that group, not your students.

For teacher educators, the study points toward pairing AI-literacy teaching with practice in self-regulation and reflection on when a chatbot starts feeling like a friend — then testing whether dependence actually falls.

What would you notice if you tried one task without asking first?

What this means for you

For you, as a teacher, professor or parent watching your own students, the study offers no proven remedy yet, only a pattern to keep in view in your own classroom. Notice whether the most confident AI users are also the quickest to skip checking, and whether a chatbot has quietly become something like a companion to them. Watch whether that confidence ever translates into less reliance, and hold the question open.

Who paid: The research was supported by the Department of Education of the Chinese Government and by a Ministry of Education Humanities and Social Sciences Research Project (Grant No. 23JDSZ3046); the article states no funder role in design, analysis or reporting, and no equipment or software was lent.

The longer versionThe passages copied from the paper, the pictures, and every source with how much of it we read · 1205 words · about 6 minRead it →Close

Most Future Teachers Reach for AI First. A Third Don't Check Its Answers.

The students who rated their own AI skills highest were not the best protected from leaning on it, a survey of 565 Chinese education majors found.

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

Among 565 students training to become teachers in China, the habits of reliance on generative AI were already the ordinary case rather than the exception. Seventy-nine percent said AI had become indispensable to their learning; 78 percent felt unaccustomed after a stretch without it; 54 percent said their first reaction to a task was to ask AI first; and 31 percent used its answers without checking them¹. The study was conducted by Bin Jiang and Zhuo Wang of the School of Educational Science at Qingdao University, with support from the Department of Education of the Chinese Government and a Ministry of Education research project, and published in Frontiers in Psychology.

The survey also asked what predicts that reliance. The strongest links were to impulsivity — a preference for the quick answer over the effortful one — and to what the authors call anthropomorphic attachment: treating the chatbot as something closer to a companion than a tool². A smaller link ran through anxiety about falling behind. Trust in AI's accuracy, on its own, did not hold up as a driver once the other factors were accounted for. And self-rated AI literacy did appear protective, but modestly, and only once the analysis separated it from the trust and peer habits that travel with it³.

The technology in question is the family of generative AI tools — ChatGPT, DeepSeek, Doubao — that in two years moved from novelty to near-ubiquity in universities, changing how students search, draft and prepare lessons⁴. The concern the authors name is cognitive offloading: delegating mental work to a machine until the habit displaces the work of thinking independently. For this particular group the stakes are doubled, because they are learners now and will be the ones modelling judgment for a generation of schoolchildren later⁵.

This is not the first study to look at what drives over-reliance. The article describes a study in which the strongest predictor of practising teachers' AI "addiction" was cognitive absorption — a kind of immersive engagement — rather than beliefs about how useful the tool was⁶. The pattern in both studies points the same way: what pulls people into dependence is not a reasoned appraisal of what AI does for them. It is how absorbing, and how companionable, the tool feels.

The five interviews in this study put flesh on that. The informant who most vividly personified the AI — "I want AI to be my classmate, my friend, my teacher" — also reported the most difficulty working without it, while the two who insisted it was "just a tool… the hands and feet, while the person is the brain" were the least reliant⁷. A separate gradient separated the heavier-relying first-years, who judged AI "rarely wrong," from critical upper-year students who caught fabricated citations and estimated only 20 to 30 percent of AI output was usable⁸.

What makes the prevalence numbers worth sitting with is that these students are not naive about it. Forty-four percent worried that long-term reliance would erode their teaching ability, and 47 percent said they wished to reduce their dependence⁹. The awareness is already there. The question the study raises is what, if anything, converts awareness into different behaviour.

The authors' own reading of their results points away from the obvious remedy. Because impulsivity was the strongest driver, they suggest the highest-yield strategy is likely to be building self-regulation — goal setting, monitoring one's own progress, tolerating the discomfort of a hard first attempt — into teacher-training coursework, so that pausing before delegating becomes a trained habit¹⁰. Because attachment also drove reliance, they propose short reflective exercises that make the pull of a human-like chatbot visible: noticing when you are treating it as a companion rather than a tool¹¹. And on literacy training, their verdict is careful. It should still be taught, but as a complement rather than a standalone safeguard, because students who feel more confident with AI also tend to trust it more and sit in peer groups that use it more — so literacy taught alone risks producing confident use without curbing dependence¹².

The limits here matter as much as the findings. The design was cross-sectional: everyone was measured at one moment, so the study cannot establish what came first, and some of these relationships are plausibly reciprocal — reliance may both follow from and deepen the sense of the chatbot as a companion¹³. Nothing was tested in the way of an actual intervention, so no one can yet say that teaching self-regulation will reduce dependence; that remains a hypothesis. The sample was recruited anonymously and by peer referral, so the authors could not control who took part: 96.1 percent were in Shandong province, 76.8 percent were women, and 60.5 percent were first-year students¹⁴. These numbers describe that group. They are not a reading of your students, your classmates or your child. And every measure was self-reported, including AI literacy — which captures how competent students believe they are, not how competent they are¹⁵.

Here is how we read it. The students most at ease with these tools may be the ones least likely to notice when they are leaning on them — not from carelessness, but because fluency makes the leaning feel like thinking. That is an expectation, not something this study measured, and it is testable: it would be wrong if students who score highest on a hands-on, marked test of AI skill turned out to be the least dependent, with their self-ratings matching their scores. Until someone runs that test, the practical move for a teacher, a parent or a student is small and unglamorous. When ease with a tool feels highest, treat that as the moment to check rather than the moment to relax: attempt the first step of a task unaided, or ask a colleague to look at one AI-assisted piece of work with fresh eyes. And when you see the student who never seems to struggle, ask what they actually verified.

The second thing we would watch for is subtler. A felt need and a real benefit can look identical from the inside. If the students who call AI indispensable can still do comparable work unaided — at the same quality, if not the same speed — then the indispensability is a habit wearing the costume of a necessity. If they cannot, it is a genuine gain and the conversation changes. The way to tell is an experiment anyone can run: do one ordinary task the old way, and notice honestly whether the result is worse or merely slower. Teachers and school leaders can build that same experiment into coursework — a first draft written without help, a self-check before the tool is opened — not as an add-on but as the assignment itself. What this study makes possible is a question to bring to the next AI policy meeting, the next lesson plan, the next time a student says they could not have done it without the machine:

Could they have? And how would anyone know?

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

Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - the article this story is about — the whole article — the passage: "Dependency markers were already widespread: 79% called AI indispensable to their learning, 78% felt unaccustomed without it, 54% “asked AI first,” and 31% used its answers without verification."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - the article this story is about — the whole article — the passage: "The strongest drivers were impulsivity (β = 0.40, p < 0.001) and anthropomorphic attachment (β = 0.28, p < 0.001), with a smaller significant contribution from AI anxiety (β = 0.15, p = 0.005), supporting H1, H2, and H4."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - the article this story is about — the whole article — the passage: "Consistent with H5, perceived AI literacy had a significant protective effect (β = −0.19, p < 0.001). This protective role was partly muted at the bivariate level (zero-order r = −0.14) because perceived literacy correlated positively with trust (r = 0.19) and social norms (r = 0.38), both associated with greater dependency."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - the article this story is about — the whole article — the passage: "The rapid diffusion of generative artificial intelligence (GenAI) tools—such as ChatGPT, DeepSeek, and Doubao—has transformed how university students search for information, draft assignments, and prepare lessons."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - the article this story is about — the whole article — the passage: "For pre-service teachers, this risk is doubled. As current learners, they are exposed to the same offloading dynamics as their peers; as future educators, the habits and judgment they form now will shape how a next generation of pupils learns to use—or over-rely on—AI."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - the article this story is about — the whole article — the passage: "A first generation of empirical studies has begun to map the antecedents of AI dependency (Zhang et al., 2024; Li and Xing, 2025), and in the teacher context. Du et al. (2026) applied the I-PACE model and found cognitive absorption to be the strongest predictor of teachers’ GenAI addiction."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - the article this story is about — the whole article — the passage: "The clearest convergence concerned anthropomorphic attachment: the informant who most vividly personified AI (“I want AI to be my classmate, my friend, my teacher”) also reported the most difficulty working without it, whereas the two who insisted AI is “just a tool… the hands and feet, while the person is the brain” were the least reliant."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - the article this story is about — the whole article — the passage: "A trust-versus-verification gradient distinguished the more reliant first-years, who judged AI “rarely wrong,” from the critical upper-years, who caught fabricated citations and deemed only 20–30% of AI output usable."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - the article this story is about — the whole article — the passage: "At the same time, a substantial minority voiced professional unease: 44% worried that long-term reliance would erode their teaching ability, and 47% wished to reduce their dependence."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - the article this story is about — the whole article — the passage: "Because impulsivity was the strongest driver, the highest-yield strategy is likely to embed metacognitive self-regulation—goal setting, progress monitoring, and the deliberate tolerance of “desirable difficulties”—into coursework (Zimmerman, 1990; Bjork et al., 2013), so that pausing to think before delegating to AI becomes a trained habit."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - the article this story is about — the whole article — the passage: "Because anthropomorphic attachment also drove reliance, programs might add brief reflective activities that make the parasocial pull of human-like agents visible—for instance, prompting students to notice when they treat a chatbot as a companion rather than a tool (Skjuve et al., 2021)."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - the article this story is about — the whole article — the passage: "Perceived AI literacy should still be taught, but as a complement rather than a stand-alone safeguard: because more self-assured students also trust AI more and are more embedded in AI-using peer norms, literacy training delivered without self-regulation support risks raising confident use without curbing dependency."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - the article this story is about — the whole article — the passage: "The design was cross-sectional and cannot establish temporal ordering or rule out reverse causality; the SEM paths are theoretically motivated associations, not demonstrated effects, and some relations are plausibly reciprocal (for example, reliance may both follow from and deepen anthropomorphic attachment)."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - the article this story is about — the whole article — the passage: "Because recruitment relied on an anonymous convenience and peer-referral strategy, we could not control the sample’s composition: it was geographically concentrated in Shandong (96.1%, chiefly Jinan and Qingdao), predominantly female (76.8%), and skewed toward first-year students (60.5%)."
Jiang, B., Wang, Z. (2026). Addicted to AI? Prevalence and psychological drivers of generative AI dependency among pre-service teachers. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1922716 - the article this story is about — the whole article — the passage: "All measures were self-reported, including perceived AI literacy, which by construction captures self-assessed rather than tested competence; future work should pair self-report with objective literacy tasks and behavioral indicators of reliance."

The findings of other studies mentioned here are known to us through this document, which is the one we read; we did not open each of those studies.