Information and education
It may help organise general information and explain complex concepts in clearer language, provided that accuracy is checked.
Education and research
Artificial intelligence and mental health: possibilities, limits and risks.
Generative artificial intelligence can organise information, support certain tasks and open new avenues for research. However, a fluent or convincing response is not the same as clinical understanding and does not guarantee accuracy.
Juan Blesa Robles is a doctoral researcher in Psychology at the University of Córdoba. His current research line addresses generative artificial intelligence in mental health.
Sessions available in English and Spanish
A starting point
A response that sounds human is still generated by a system.
Generative AI learns patterns from large datasets to produce text, images or other content. Conversational models can maintain plausible dialogue, but this does not mean that they understand a personal history in a clinical sense or can assume professional responsibility.
Many general-purpose tools are used for emotional support even though they were not specifically designed or evaluated for mental health. WHO has highlighted the need to study their effects, include appropriate oversight, protect vulnerable people and keep human well-being at the centre.
Possibilities that require evaluation
AI may support specific tasks without becoming a therapist.
Administrative tasks
It may reduce work involved in documentation or repetitive processes when use is clearly defined, data are protected and a person reviews the output.
Training
It may generate examples or simulations for learning, but materials need review because models can invent details or oversimplify.
Research
It may help organise, summarise or explore scientific information without replacing methodology, source verification, peer review or researcher judgement.
Risks worth understanding
A convincing response is not proof of reliability.
False information
Models may invent facts, references or explanations and present them confidently. This is often called model confabulation or hallucination.
Bias and inequality
Systems may reproduce bias in their data and perform differently across languages, cultures or populations.
Privacy
Information entered into a chatbot does not necessarily receive the same safeguards as data handled within healthcare.
Lack of clinical context
A model does not independently conduct a comprehensive assessment of history, functioning, risk, non-verbal communication and personal circumstances.
Over-reliance
People may attribute understanding, authority or abilities to a system that it does not possess and rely too heavily on its responses.
Emotional dependence
Anthropomorphising a chatbot may create confusion about its nature or an attachment that displaces human relationships and support.
Inadequate crisis response
A general-purpose chatbot is not an emergency service, cannot verify a person’s safety and cannot guarantee an appropriate referral.
Unclear accountability
A tool may generate a recommendation, but it does not assume the clinical, ethical and legal responsibility of a healthcare professional.
Using AI with care
Some precautions can reduce risk, although they cannot remove it.
Information, not diagnosis
Use these tools for general questions, not to confirm or rule out anxiety, depression or another psychological difficulty.
Verify before deciding
Check important claims using official sources or a qualified professional, particularly when they affect your health.
Protect sensitive data
Avoid entering names, clinical records, documents or information about other people, and review the provider’s privacy terms.
Maintain human oversight
Do not change medication, treatment or important decisions solely on the basis of an automated response.
Take additional care with minors
Teenagers may attribute human qualities to these systems or use them when vulnerable; support from adults and professionals remains essential.
Ethics and governance
In health, innovation also means protecting autonomy, safety and equity.
WHO proposes protecting autonomy, promoting well-being and safety, ensuring transparency, maintaining human accountability, supporting inclusion and continuously evaluating AI systems used in health.
In the European Union, AI regulation follows a risk-based approach and sets different obligations depending on intended use. This reinforces a basic principle: a general productivity tool should not automatically be treated as a validated clinical tool.
Juan Blesa’s doctoral research line
Research before making promises.
Juan is a doctoral researcher in Psychology at the University of Córdoba. His ongoing research line examines generative artificial intelligence in mental health.
Being a doctoral researcher does not mean that findings are already available or turn a general-purpose tool into a treatment. This page will be updated when public evidence can be communicated rigorously; until then, it makes no claim to original findings, publications or a proprietary method.
Care at Pulso en Cauce remains a relationship between a person and a professional. Sessions are available in person in Córdoba or online, in Spanish or English.
Frequently asked questions
Frequently asked questions about AI and mental health
Clear answers to help distinguish information, emotional support and professional psychological care.
01Can artificial intelligence replace a psychologist?
No. It may provide information or support specific tasks, but it cannot replace clinical assessment, a therapeutic relationship, contextual knowledge or the responsibility of a healthcare professional.
02Can ChatGPT diagnose anxiety or depression?
It should not be used to establish or rule out a diagnosis. Symptoms may have different causes and need to be assessed in terms of duration, intensity, impact, risk and personal context.
03Is talking to a chatbot confidential in the same way as talking to a psychologist?
Not necessarily. Professional confidentiality and the safeguards applied to healthcare data are not equivalent to a technology provider’s privacy terms. Avoid sharing identifying information or clinical documents.
04Is AI always wrong?
No, but it can present correct and incorrect responses with the same confident tone. Confidence does not demonstrate reliability, so important claims should be verified.
05Can AI be used to support therapy?
Some applications may support specific tasks when they have a defined purpose, sufficient evidence, data safeguards, human oversight and ethical protections. This is different from replacing therapy with a general-purpose chatbot. Pulso en Cauce does not offer AI-based therapy.
06What is Juan Blesa researching in this area?
His doctoral research line at the University of Córdoba addresses generative artificial intelligence in mental health. It is ongoing research: this page does not present original results, publications or conclusions that do not yet exist.
07Should a teenager use AI for emotional support?
A general-purpose tool cannot replace support from trusted adults or professional care. Additional caution is needed with minors because of privacy, inaccurate information, emotional dependence and inadequate responses to risk.
08Can I receive psychological support in English?
Yes. Pulso en Cauce offers in-person sessions in Córdoba and online sessions in Spanish and English. Care is provided by a professional, not by an artificial intelligence system.
Sources and further reading
Sources and further reading
References used to prepare and review this information.
- World Health Organization. WHO calls for safe and ethical AI for health
- World Health Organization. Ethics and governance of AI for health: guidance on large multimodal models
- World Health Organization. Towards responsible AI for mental health and well-being
- National Institute of Standards and Technology. Artificial Intelligence Risk Management Framework: Generative AI Profile
- European Union. European Union Artificial Intelligence Act
- Spanish Ministry of Health. 024 support line for suicidal behaviour
Pulso en Cauce
Finding direction can begin with a conversation.
The first session is a brief, free and no-obligation meeting to get to know each other, answer questions and consider whether this space feels right for you.