“I have been thinking of going into therapy for a while, but I also have a lot of doubts about that. Can I try AI chatbots, such as ChatGPT, Gemini or Claude as a therapist, just to try if that works? Or is it better to go to a therapist after all?”
The short answer
At this moment, AI agents are certainly not an effective replacement for care from a mental health professional. Professionals have a license to provide qualified and safe care. AI agents can provide inaccurate and in some cases dangerous advice and cannot provide diagnoses for mental health conditions. This means that the short answer to this question from a patient is: no, AI is not a good alternative for going to a professional clinician for help with mental health problems.
Especially in a crisis situation, it is important not to rely on AI, but to seek immediate human help.
What the evidence shows so far
Although there is an increasing number of studies examining the effects of using AI as a therapist for mental health problems, it is not yet known whether these interventions are effective. The first well-designed studies do show that AI, under experimental conditions (so with modifications specifically for the study) can have effects on mental health problems. That is not surprising, because there is already a long tradition of digital interventions for mental health problems. These digital interventions are based on self-help, meaning that users learn how to apply a psychological treatment to themselves. There are two broad types of such digital treatments: self-guided treatments, without any human support, and guided self-help, meaning that a human coach or clinician helps the patient work through the treatment. Both are effective when compared to control groups, but guided self-help is substantially more effective than self-guided treatments, and is as effective as face-to-face therapies in several mental health conditions. Currently it is not known whether AI therapies (without involvement of clinicians) are as effective as self-guided treatments, or if they can be as effective as guided self-help and face-to-face therapies.
AI and the accessibility of care
If some AI therapies would indeed turn out to be as effective as face-to-face therapies, that would mean that many more patients worldwide could be provided with effective treatments. That is important, because many patients currently do not have access to face-to-face therapies. It is, however, also very much possible that AI therapies are not more effective than self-guided treatments. This would still mean that more people will get access to evidence-based treatments, but in that case AI therapies may not add very much to what is already existing. However, they may make treatments more accessible or attractive for some people.
All treatments have limits
It must be kept in mind, however, that all treatments have limited effects (on average, a 50% symptom reduction is realized in 30 to 42% of patients with a common mental disorder). Even if AI therapies can approach these numbers, there will still be many patients who will not be helped sufficiently with these treatments, and they will need additional treatment.
Using AI safely
The American Psychological Association has released an excellent guide for patients on AI agents and how to use them safely. If you are considering to use AI, you should certainly read this guide. It summarizes what you should know when you use AI, for example that such chatbots always tends to agree with you, that they are designed to keep you engaged, sound confident even when they are wrong, and that they can feel like it “knowing” you, while in reality they do not. It is also important to keep in mind most chatbots are provided by for-profit companies that stores your personal information. If you use such conversational agents, for example for organizing your thoughts and reflections, it is important to realize that mental health information from AI can be wrong, and that it cannot give diagnoses. And if you go to a clinician, be open about your use of AI.
Pim Cuijpers is professor emeritus of clinical psychology and scientific director of Metapsy. He has been involved in more than 1,100 scientific studies, mostly on psychological treatments of mental health problems. This is one of a series of evidence summaries in which Prof. Cuijpers tries to answer questions from patients and clinicians, based on what is known in science about treatments. The knowledge is mostly drawn from collective work of the Metapsy collaboration of at least 15 years. Do you have other questions you would like Prof. Cuijpers to discuss? Feel free to contact us.
Literature
- American Psychological Association. APA Guide to Navigating AI-Generated Advice Thoughtfully and Safely. apa.org/topics/artificial-intelligence-machine-learning/guide-navigating-ai
- Karyotaki E, Efthimiou O, Miguel C, Bermpohl FMG, Furukawa TA, Cuijpers P; Individual Patient Data Meta-Analyses for Depression (IPDMA-DE). (2021). Internet-based cognitive behavioral therapy for depression: A systematic review and individual patient data network meta-analysis. JAMA Psychiatry, 78(4), 361–371.
- Tong L, Panagiotopoulou OM, Cuijpers P, Karyotaki E. (2024). The effectiveness of self-guided interventions in adults with depressive symptoms: A systematic review and meta-analysis. EBioMedicine, 105, 105208.
- Cuijpers P, Noma H, Karyotaki E, Cipriani A, Furukawa TA. (2019). Effectiveness and acceptability of cognitive behavior therapy delivery formats in adults with depression: A network meta-analysis. JAMA Psychiatry, 76(7), 700–707.