Why you shouldn't trust chatbots with your problems: AI risks in psychology

AI-powered chatbots are increasingly being used as "psychologists," promising accessible and anonymous help. However, this practice carries serious risks, warns Ekaterina Orlova, psychologist and Deputy Director of the Institute of Clinical Psychology and Social Work at Pirogov University. In an interview with Gazeta.Ru, she explained why confiding your deepest concerns to algorithms can be dangerous and how it may cause harm. Let's examine the key problems of using AI in psychological support and why human specialists remain irreplaceable.

1. Vulnerability of Personal Data
   One of the main threats is the risk of confidential information leaking. Sharing your fears, traumas, or personal experiences with a chatbot is essentially entrusting them to a third-party system that may be vulnerable.

• How this happens:
• Data may fall into the hands of hackers due to cyberattacks
• Unscrupulous companies may use the information for advertising, manipulation, or resale
• Even anonymized data can sometimes be deanonymized, threatening your privacy
• Consequences:
• Blackmail or discrimination based on personal information
• Loss of trust in digital services

Example: In 2023, a data leak from a popular AI service exposed user conversations, including their psychological complaints.

Tip: If you're using a chatbot, avoid sharing personal details (name, address, specific dates) and review the service's privacy policy.

2. Lack of Empathy and Nonverbal Analysis
   AI cannot pick up on emotional nuances that are critical in psychology. A human therapist evaluates not only words but also tone, facial expressions, gestures, and pauses. Chatbots are incapable of this.

• Problems:
• AI does not detect hidden despair or anxiety, which may lead to incorrect advice
• Algorithms do not empathize, making users feel misunderstood
• In critical situations (such as suicidal thoughts), AI may give dangerous advice or fail to notice severity
• Consequences:
• Worsening condition due to inadequate support
• Missed warning signs requiring urgent help

Example: A chatbot may respond with a template reply to the sarcastic phrase "everything's great, I want to disappear," ignoring the red flag.

Tip: For serious emotional issues, consult licensed psychologists who can evaluate your condition holistically.

3. Algorithmic Bias
   Neural networks are trained on human-generated data and may reproduce stereotypes or discriminatory approaches embedded in that data.

• Manifestations:
• AI may give unsuitable advice to specific groups (e.g., ignoring cultural nuances)
• Algorithms may reinforce biases related to gender, age, or mental conditions
• A limited dataset leads to generic and often unhelpful recommendations
• Consequences:
• Incorrect advice that fails to consider individuality
• Reinforcement of stigma surrounding mental health

Example: A neural network trained on Western data may misinterpret user complaints from a different cultural background, labeling them as "not serious."

Tip: If AI gives strange or inappropriate advice, do not follow it without consulting a professional.

4. Illusion of Replacing Professionals
   Chatbots create the false impression that they can fully replace a psychologist. This is especially dangerous for people with serious conditions such as depression or PTSD.

• Why this is harmful:
• Users delay seeing a specialist, relying on AI
• Chatbots cannot develop a long-term therapy plan or work through deep trauma
• Lack of human contact reduces the effectiveness of support
• Consequences:
• Lost time in treating serious conditions
• Dependence on AI instead of seeking real help

Example: Someone with anxiety disorder may receive breathing exercise tips from a chatbot, but without professional therapy, the issue remains unresolved.

Tip: Use AI as temporary support (e.g., for relaxation), but for serious problems, book an appointment with a psychologist.

5. Lack of Transparency and Accountability
   Chatbots function as a "black box": users don't know how their data is processed or who is responsible for errors.

• Problems:
• It's unclear where your messages go or how they’re stored
• In case of harm (e.g., due to bad advice), there is no clear responsible party
• Legal regulations for AI-psychologists are not yet established
• Consequences:
• Lack of user rights protection
• Risk of trust abuse

Example: If a chatbot suggests the wrong way to cope with a panic attack, the user cannot file a complaint like they would with a licensed professional.

Tip: Choose services with transparent data policies and avoid platforms that don’t explain how your information is used.

How to Use Chatbots Safely
• Limit personal data: Don’t share information that could identify you
• Verify the source: Use reputable platforms (e.g., Grok 3 by xAI)
• Combine with professional help: Chatbots are suitable for light issues (e.g., stress tips), but not for therapy
• Be critical: Don’t take AI advice as truth, especially in crises
• See a specialist: For serious issues, find a psychologist through platforms like Yasno or Alter

Alternatives to Chatbots
• Online therapy: Platforms like Yasno, Zigmund.Online, or Alter offer sessions with licensed psychologists starting from 2000 ₽ per session
• Hotlines: Free psychological help services (e.g., 8-800-2000-122 in Russia)
• Self-help: Meditation, emotion journals, or psychology books (e.g., Emotional Intelligence by Daniel Goleman)
• Support groups: Interest- or topic-based communities (e.g., anxiety, depression) on Telegram or VK

Conclusion
AI chatbots may seem like a convenient solution for psychological support, but their use comes with risks: data leaks, lack of empathy, algorithmic bias, the illusion of professional replacement, and lack of transparency. As psychologist Ekaterina Orlova notes, the best approach is to combine technology with human specialists. In 2025, AI can be a helpful tool for minor issues, but for serious problems, trust licensed professionals.