THEor you can disturb every time you want And also in the middle of the night, that is, just when things are worse. NHe never interrupts and always finds the right words. Does not judge and costs very littlemaybe nothing. Is a psychologist in flesh and blood does the same, at the same price? There is not a lot of investigation to explain The success and diffusion of chatbots and AI for psychological and emotional support. A real boom, how The Harvard Business Review, according to which the personal support theme (between the therapy/company voices, organize my life and find a purpose) has far become the first use of artificial intelligence in 2025, stealing ground from technical assistance.
Other studies confirm it, especially among young people: Fin from their appearance, at the end of the decade 2010-2021, Therapeutic robots have constantly gained ground, and patients. How do they work (since, yes, of course, to some extent they work)?
The artificial intelligence instead of a psychologist in the flesh and blood? Pros and cons of therapy chatbot
We talked about it with a psychologist and psychotherapist expert in technology, curious about the theme and, above all, realist. “Just take a ride on Reddit for understand how much the phenomenon is a reality“, explains Tommaso CiulliSecretary of the Italian Society of Digital Psychology and Councilor Order Psychologists of Tuscany. “Many people compare the therapy they can do with a chatbot daily with the one they did or they still do with a professional in flesh and blood and explain Why and when the AI is better».
Beyond the economic factoron which there is no game, they say for example that the AI is “better” because he doesn’t judge. But in reality not even a therapist should. They say that It is better because “if I speak of X, Yez, the therapist considers only ZAi takes all the themes. ” «Any psychologist should find these very interesting comments and reflect us. As professionals we can continue to stay closed in our castles by repeating that we are better And repeating that that of AI is a simulation of empathy. Or, finally, to accept that people use it and ask us ».
Most interactions with psychological support apps take place at night (Getty Images)
The empathy report is illusory (but it works)
Ciulli has a certain experience of psychological chatbots: in 2018 he collaborated on a European project, Teo, who had clinically tested a psychological support AI intended for therapists. “It was a generative AI capable of supporting the professional, for example in the moments when there is not physically. In general, The results were very positive, especially for the management of anxiety problems, depression and low motivation levels». What is happening today is obviously different: it is the patient alone to turn to artificial intelligence: only with his discomfort, more or less important. What establishes with the machine is an illusory empathy relationship But, surprise, it could work quite well.
Brief history of empathic artificial intelligence
Ciulli explains: «Since, in 2017, the transformer architecture was introduced, the AI has been able to understand the global context of a text. It is on this ability that the explosion of the AI is grafted, Suddenly able to provide the human interlocutor to the meaning of meaning. Suddenly able to appear empathic». So empathetic as to become lovable, and even potential objects of a romantic love, such as explains this 2022 study.
Psychobot: the two artificial intelligence models
From generic empathy to psychological support, however, it passes there. To date There are essentially two models of “psychobot”, As Ciulli explains: on the one hand those trained by psychologists and psychotherapists with notions of therapy cognitive-behavioral (TCC). On the other the generative ones.
Le Ai Attrastrate by professionals
The former substantially guide users a manage emotions through structured techniques (of relaxation, for example): The therapeutic approach is quite rigid. To the point that, as soon as the patient steps on the side in describing how he feels or in the articulate his own thoughts, The bot is programmed to redirect it to the default path of the clinical tools he knows. It is a tool that is based on rules written by experts, therefore safe, but with obvious limits. This category is part, for example, WYSAsupported and offered by the United Kingdom National Health Service to make up for the deficiencies of the public service. But also Weabot, among the first mental health supports, founded in 2017 by a psychologist graduated from Stanford, who has as soon as the closing of the app is announced.
To the generative
The generative, however, not only include the context but generate a completely new and original content. It is Bot relational or conversational sayings that are based on advanced linguistic models (the so -called Large Language Models) to create interactions that seem surprisingly real. “I’m a lot condescendingand this might seem like a pro. But let’s say that the “patient” says: “I am unsatisfied, all colleagues have it with me”. Ai could answer: “I can understand you, maybe you’re right” ».
Are part of this category Chatgpt e Replythe most specific Youper And Pi. And also Therabot, the only one to the therapeutics that was subjected to a clinical study (Here the New York Times talks about it). Clinical study in which, between Ai and patients, a very high level of “therapeutic alliance” Similar to what is found in interactions with meat and blood therapists. The interactions with the software recorded peaks in moments typically associated with malaise, as in the middle of the night.
An condescending psychologist (too much?): The risk of addiction
In short, the AI has diagnostic capacity (and this is already incredible). Not only that: programmed to reassure and support, it can help reduce less serious mental and emotional problems, such as anxiety and a sense of solitude. Some doubts remain, of course: This study, published by researchers from Mit Media Lab and Openai (who developed chatgpt!), The comfort it gives could be such that people end up isolating each other even more.
The main doubts, however, concern the quality of the therapeutic response in the medium and long term, beyond the temporary satisfaction of the user. Last year, the mother of a 14 year old boy of Florida sued the chatbot Character.ai manufacturer, claiming it was responsible for his suicide.
Apart from the extreme results, the question also concerns the least serious inconvenience, and it is “The condescence: it is often too much. This, combined with total accessibility, can create addiction », continues Ciulli.
“But three reasons why people prefer chatbots to people are these elements. If a good professional, even at the cost of producing frustration in the patient, leads him to deviate from his canonical mental processes, precisely for therapeutic purposes, Artificial intelligence, on the contrary, his patient comforts him, reassures him, caresses him: always».
The problem of psychological support AI
Then, if the trained apps may appear rigid, unable to answer questions that exist those provided for by AI instructors, the generative ones are potentially explosive. AND The Tnote note of the generative IA chatbots to “hallucinated”, or to invent information. And the problem is that often it is not clear where the problem is: unlike the first type, these “have a very low transparency: it is difficult to understand what has not worked and when”.
Then there are other critical issues that concern the possible use of the AI in support of the therapist’s work. «As professionals, we could use it for example to prepare questionnaires, but also to submit them to the notes of a patient for one Sentiment analysis. Here, in this case, would be fundamental anonymize the material. Penalty the violation of ethical and privacy rules, of course, but not only. We must be aware that that AI material is a training “.
The “added value” of the professional in flesh and blood
Ciulli then reports that there is a lot of difference between pushing a patient to use a chatbot for psychological support and, instead, take note of the fact that, independently, he did it. “If the hypochondriacs have disappeared from our studies, it is because they find all the answers on Google. Those suffering from other inconveniences could also disappear». Not without consequences.
For the same category it is a complicated moment, of precise focus of their added value, as people, compared to the machines. “As he said with a provocative joke from a colleague during a conference, do we have to understand it: when will we be better than artificial intelligence?”.
I woman © RESERVED REPRODUCTION