Author – Katerina Drakos
Trigger Warning/Disclaimer: This blog post mentions suicide. If you or someone you know is experiencing suicidal thoughts or a crisis, please reach out immediately for help. A hotline in your country can be found on befrienders.org.
OpenAI recently released a new product designed to support medical care: ChatGPT Health. They describe this innovation as an “experience that securely brings your health information and ChatGPT’s intelligence together, to help you feel more informed, prepared and confident navigating your health” (Introducing ChatGPT Health, 2026). They claim that health is already one of the most common ways people use ChatGPT, and that this platform will increase users’ knowledge, preparation and confidence in managing their health (Introducing ChatGPT Health, 2026). However, while AI-powered health tools may reduce information overload and improve accessibility, their integration into help-seeking processes raises significant ethical questions regarding misinformation, autonomy and bias.
The practice of virtually navigating health information has been termed “online help-seeking” amongst the health community and is not a novel concept (Pretorius et al., 2019). With the rise of the internet, particularly in digitalised societies, people are increasingly turning to digital resources with a variety of health-related questions. In 2022, the European Commission determined that across Europe over half the population turned to the internet for health information and finding care. In Denmark 75% of people aged 16-74 years searched online for health-related topics, one of the highest percentages in Europe (Eurostat, 2022).
Online Help-seeking
Key benefits of online help-seeking include anonymity, privacy, immediacy, ease of access, inclusivity, the ability to connect with others and share experiences, and a greater sense of control over the journey. The facilitated access to information has the potential to act as a gateway for further help-seeking activities (Kauer et al., 2014; Pretorius et al., 2019). Even in suicidal individuals, which are usually the most reluctant, the internet has altered the youth suicide-related help-seeking paradigm (Seward & Harris, 2016).
However, there are some risks associated with online help-seeking. In one study based in Hong Kong of 1214 participants, individuals who search for mental health help online tend to report poorer mental health and higher suicidality (Wong et al., 2021). Another study concerning cancer patients noted that they were “overwhelmed, stressed, anxious and depressed, not only by the quantity and the complexity of the information available, but also by the fact that often online information conflicted with one another and was not tailored to the individual” (Ferraris et al., 2023). In addition to this, the legitimacy and reliability of health information online can be called into question such that a high level of digital literacy is required from the outset to discriminate between trustworthy information and misinformation (Ferraris et al., 2023).
Another potential area of caution is the risk of contributing to cyberchondria, by providing unlimited access to information. Cyberchondria is the excessive and repetitive searching for health information online (Starcevic et al., 2021). People seeking reassurance about their health may spend much of their time searching for information and attempting to determine the validity of health-related information (McMullan et al., 2019). This process contributes to the cycle in which repeated online searches increase distress and anxiety (Starcevic et al., 2021). Therefore, even though digital health literacy has been established in research as a determinant of health (López et al., 2023) , a high level of literacy could also contribute to compulsive online health search behaviour (Babu, 2025). Here, education regarding the use of technology and awareness about cyberchondria could be helpful in reducing health-related anxiety (Kalantari et al., 2025).
Artificial Intelligence (AI) for Online Help-Seeking
Given these challenges, the question arises whether AI could mitigate some of the structural weaknesses of online help-seeking. By providing reliable, evidence-based and trustworthy mental health information in a synthesised and direct manner, AI has the potential to intervene positively in this context. By having a centralised provider of trustworthy sources, users have access to legitimate mental health information and resources, thereby increasing their mental health literacy and improving the identification of problematic situations. This in turn can facilitate their help-seeking behaviour. It can also respond to misinformation and disinformation that is often seen online or on social media. While online help-seeking is characterised by ambiguity of online health information (in that it can be inaccurate, misleading or incomplete), a mental-health help-seeking chatbot powered by a RAG model (See previous blog) could ensure that the information is factual and valid (McMullan et al., 2019). This could reduce the time spent determining the validity of the information and the time spent online.
In the report “AI as a Healthcare Ally: How Americans Are Navigating the System With ChatGPT” released in 2026, OpenAI states that in general three in five adults in the US say they have used AI tools for their health or healthcare in the past three months (OpenAI, 2026). This is predominantly to check and explore symptoms, organise information, translate jargon or learn about treatment options. If people are already using AI for this purpose, specific purpose-built chatbots in this field could be extremely relevant at a population level.
Yet precisely because AI systems such as ChatGPT Health promise to centralise, filter and synthesise health information, they introduce a new layer of ethical concerns. If an AI tool is responsible for curating and selecting the information, prioritising certain sources and formulating responses in a confident, authoritative tone, the question is no longer only whether the information is accurate, but whose values are embedded in the system and how these values shape the user’s online help-seeking trajectory. If chatbot developers and designers are more concerned with engagement, other aspects of the provision of information will be forfeited, and we have no way of knowing which ones.
This raises important questions about how the model is trained, which data it relies on and according to which assumptions about health, wellbeing and appropriate care pathways it operates. Even when powered by techniques such as RAG, the selection of “reliable” sources is itself a choice of value prioritisation. Decisions about what counts as trustworthy evidence, which guidelines are prioritised, and how risk is framed inevitably reflect biases of the company who designed, developed and deployed the chatbot, such as institutional, cultural and commercial perspectives.
In January 2026 Google was criticised for displaying AI summaries that appeared at the top of search results with inaccurate health information, putting people at risk of harm (Gregory, 2026). From a design perspective, highlighting one paragraph of information does reduce the burden of online-help seeking, but if done incorrectly risks the dissemination of misinformation and disinformation. This feature of AI incorporated into the Google search engine could even potentially lead to people avoiding seeking help and having the contrary effect to what is desired when promoting accessibility to health information online. This synthesised summary should not be taken as the absolute truth, but displaying it front and centre could lead the user to accept it as such, especially if they already struggle to access trusted health information. It is therefore essential that any AI-based information provider signposts robust, researched health information from trusted sources.
The next major concern regards the ethical principle of autonomy. On the one hand, reducing the overwhelming quantity of online information and offering a synthesised, evidence-based answer may protect users from misinformation and overwhelming decision-making. On the other hand, by pre-structuring the information landscape, AI may narrow the range of perspectives presented, subtly nudging users toward specific interpretations. In mental health contexts especially, where experiences are deeply personal and culturally dependent, the risk of over-standardisation becomes apparent. Does the system enhance autonomy by clarifying options and promoting user confidence or does it constrain autonomy by determining which options are made visible in the first place?
The second issue concerns personalisation and data. For AI to provide tailored guidance, it must rely on training data and, potentially, user-provided information. However, training datasets may underrepresent certain populations (Poole-Dayan et al., 2025), reinforcing existing disparities in mental healthcare access and recognition. Moreover, personalisation mechanisms can amplify biases if they infer user needs based on incomplete or stereotypical patterns. Transparency about how data are used, how recommendations are generated and where the system’s limits lie is therefore central to maintaining trust.
Value-alignment for Help-seeking
Value alignment could be beneficial in this situation, and any technology using AI designed for mental health help-seeking should be intentionally aligned with ethical principles that guide any health professional in the field: beneficence, non-maleficence, autonomy, justice. Rather than replacing user choice, they should support it, making pathways clearer while preserving the individual’s capacity to question, compare and ultimately decide. In this sense, the challenge is not merely technical, but a human-computer, interaction design principle as well. Our goal should be to ensure that AI does not redefine what it means to seek help but instead supports the user on their unique help-seeking journey.
Overall, it is important that the population is aware of what mental wellbeing and mental health problems look like, what they can look out for, and where they should go in case of concerns. AI could be present throughout this help-seeking journey and can facilitate this search in a natural and conscious manner. Through a human and computer partnership, AI could increase confidence and assist the population in navigating complex systems, such as mental healthcare services. AI systems for health-related, help-seeking must therefore be intentionally designed to uphold ethical principles, ensuring that technological efficiency does not override user autonomy, equity and informed decision-making.
References
Eurostat. (2022). EU citizens: Over half seek health information online [Dataset]. https://ec.europa.eu/eurostat/web/products-eurostat-news/-/edn-20220406-1.
Ferraris, G., Monzani, D., Coppini, V., Conti, L., Maria Pizzoli, S. F., Grasso, R., & Pravettoni, G. (2023). Barriers to and facilitators of online health information-seeking behaviours among cancer patients: A systematic review. DIGITAL HEALTH, 9, 20552076231210663. https://doi.org/10.1177/20552076231210663.
Gregory, A. (2026, January 11). ‘Dangerous and alarming’: Google removes some of its AI summaries after users’ health put at risk. The Guardian. https://www.theguardian.com/technology/2026/jan/11/google-ai-overviews-health-guardian-investigation.
Introducing ChatGPT Health. (2026, July 1). https://openai.com/index/introducing-chatgpt-health/.
Kalantari, A., Valizadeh-Haghi, S., Starcevic, V., Shahbodaghi, A., Rahmatizadeh, S., Zayeri, F., & Khazaal, Y. (2025). The relationship between e-Health literacy and cyberchondria in Iranian students of health sciences. Frontiers in Psychiatry, 15, 1421391. https://doi.org/10.3389/fpsyt.2024.1421391.
Kauer, S. D., Mangan, C., & Sanci, L. (2014). Do Online Mental Health Services Improve Help-Seeking for Young People? A Systematic Review. Journal of Medical Internet Research, 16(3), e66. https://doi.org/10.2196/jmir.3103.
McMullan, R. D., Berle, D., Arnáez, S., & Starcevic, V. (2019a). The relationships between health anxiety, online health information seeking, and cyberchondria: Systematic review and meta-analysis. Journal of Affective Disorders, 245, 270–278. https://doi.org/10.1016/j.jad.2018.11.037.
McMullan, R. D., Berle, D., Arnáez, S., & Starcevic, V. (2019b). The relationships between health anxiety, online health information seeking, and cyberchondria: Systematic review and meta-analysis. Journal of Affective Disorders, 245, 270–278. https://doi.org/10.1016/j.jad.2018.11.037.
OpenAI. (2026). AI as a Healthcare Ally: How Americans are navigating the system with ChatGPT.
Poole-Dayan, E., Roy, D., & Kabbara, J. (2025). LLM Targeted Underperformance Disproportionately Impacts Vulnerable Users (arXiv:2406.17737). arXiv. https://doi.org/10.48550/arXiv.2406.17737.
Pretorius, C., Chambers, D., & Coyle, D. (2019). Young People’s Online Help-Seeking and Mental Health Difficulties: Systematic Narrative Review. Journal of Medical Internet Research, 21(11), e13873. https://doi.org/10.2196/13873.
Seward, A.-L., & Harris, K. M. (2016). Offline Versus Online Suicide-Related Help Seeking: Changing Domains, Changing Paradigms. Journal of Clinical Psychology, 72(6), 606–620. https://doi.org/10.1002/jclp.22282.
Starcevic, V., Schimmenti, A., Billieux, J., & Berle, D. (2021). Cyberchondria in the time of the COVID‐19 pandemic. Human Behavior and Emerging Technologies, 3(1), 53–62. https://doi.org/10.1002/hbe2.233.
Wong, K., Chan, C. S., Chan, M., Wong, C., Cheng, Q., Xiong, C., & Yip, P. (2021). Who seeks help online? Comparing online and offline help-seeking preferences amongst youths with suicidal ideation. Journal of Affective Disorders, 292, 21–29. https://doi.org/10.1016/j.jad.2021.05.056.
Further Reading/Watching/Listening:
Podcast: Freakonomics, Can AI save your life? https://freakonomics.com/podcast/can-a-i-save-your-life/.
Image Attribution
Generated by: Better Images of AI
Date: 2025