The benefits and future potential of generative artificial intelligence (GAI) on mental health: a Delphi study

“Fifteen experts with professional backgrounds in both technology and mental health participated, contributing diverse and specialised insights into the role of GAI in mental health practice (Niederberger & Spranger, 2020).”

“…rather than as replacements, especially for individuals with severe or complex mental health conditions and for those living in areas with limited access to qualified professionals (Blease & Torous, 2023; Essien & Asamoah, 2020; Miner et al., 2020; Patel et al., 2007; Torous & Blease, 2024).”

“Fifteen mental health professionals were selected through purposive sampling to ensure a well-informed and contextually relevant expert panel (Beiderbeck et al., 2021).”

“These insights align with frameworks in learning health systems and precision public health that leverage data-driven approaches to improve intervention design (Elyoseph et al., 2023; Friedman et al., 2015; Xian, 2024).”

“A key advantage of naturalistic chatbots is their accessibility (Abd-Alrazaq et al., 2020).”

“For example, AI-powered chatbots and conversational agents can deliver evidence-based relaxation techniques and therapeutic dialogues that help users manage anxiety and depression (Abd-Alrazaq et al., 2020; Sharma et al., 2023).”

“Recent studies emphasise that transparency, fairness, and accountability are essential for trustworthy AI systems that protect patient autonomy and privacy while maintaining clinician–patient trust (Cheong, 2024; Ott & Dabrock, 2022).”

“Data were organised and coded manually using Microsoft Excel as a practical tool for sorting, labelling, and grouping responses (Lochmiller, 2021).”

“By using natural language processing and machine learning, these tools can improve emotional regulation and user engagement in mental health self-care (Elyoseph et al., 2023; Liu et al., 2023).”

“These findings align with research indicating that AI can improve diagnostic accuracy and streamline administrative processes, enabling clinicians to devote more attention to direct patient care (Rollwage et al., 2024; Sendak et al., 2020).”

“Recent research consistently confirms that PU remains one of the strongest predictors of technology adoption across diverse domains (Ma et al., 2025).”

“As emphasised in previous studies, ethically governed personalisation can strengthen therapeutic alliance without compromising user autonomy (Blease & Torous, 2023; Yu et al., 2023).”

“It was anticipated that the attendance of experts would provide valuable professional expertise and insights, significantly advancing this study (Spranger et al., 2022).”

“This vision aligns with broader healthcare innovations that employ machine learning to refine prediction accuracy and personalise care pathways (Al-Worafi et al., 2023; Pham et al., 2022).”

“In mental health care, where confidentiality and empathy are vital, building transparent and ethically governed AI systems is a prerequisite for clinical acceptance and long-term integration into professional practice (Saeidnia et al., 2024; Svensson et al., 2025).”

“However, researchers have also raised concerns about potential limitations, including algorithmic bias, the risk of overreliance on AI-generated empathy, and the lack of human accountability in automated interactions (Blease & Torous, 2023; Hitch, 2023).”

“Scholars have called for clear guidelines to ensure that AI enhances rather than replaces human expertise and clinical judgement (Cheng et al., 2023; Doraiswamy et al., 2020).”

“Previous research has similarly underscored that conversational AI can deliver empa-thetic, language-based interactions that facilitate early symptom identification and emotional regulation (Balcombe, 2023; Ettman & Galea, 2023).”

“Studies have shown that usability and accessibility significantly influence user trust and intention to use a system (Al-Abdullatif, 2024).”

“Attitude toward use reflects an individual’s overall positive or negative feeling toward using a technology and is a direct predictor of behavioural intention (Ovsyannikova et al., 2025).”

“The interpretation of the findings relies heavily on the expertise and judgement of the participating professionals (Schmalz et al., 2021).”

“This analytical process enabled categorising expert inputs into coherent thematic dimensions, which subsequently informed the structure of the second-round questionnaire (Deveci et al., 2020).”

“For instance, OpenAI introduced the Chat Generative Pre-training Transformer (ChatGPT) on November 30, 2022, as a highly capable conversational system developed in San Francisco (Singh, 2023).”

“This perception underscores the importance of trust, a critical extension of the Technology Acceptance Model in mental health contexts (Blease & Torous, 2023).”

“Empirical studies in healthcare and AI adoption consistently find that trust mediates between user intention and acceptance, while perceived risk negatively influences acceptance (Goh et al., 2024; Kauttonen et al., 2025).”

“They were required to assign distinct rankings without duplication to ensure clarity and to facilitate statistical analysis (Boel et al., 2021).”

“Trained on extensive textual data, GAI can produce written responses that resemble human language and thought patterns (Naidu & Sevnarayan, 2023).”

“By supporting administrative functions and providing structured assistance based on evidence-informed therapeutic principles, GAI can reduce workload pressures and enhance efficiency (Hamdoun et al., 2023).”

“In the field of healthcare, AI has been widely applied to support evaluation activities, such as algorithmic data processing, and to drive innovations in robotics, smart devices, and automated medication delivery systems (Al-Worafi et al., 2023; Woodnutt et al., 2024).”

“Users who perceive AI systems as empathetic, reliable, and beneficial are significantly more willing to continue using them (Shen et al., 2024).”

“Studies show that intuitive and emotionally attuned conversational AI systems, such as hybrid chatbots, enhance engagement, reduce communication anxiety, and foster a sense of therapeutic presence (Wah, 2025).”

“Strengthening transparency, accountability, and cultural sensitivity can reinforce trust, shape positive professional attitudes, and ultimately support the responsible global adoption of GAI systems (Blease & Torous, 2023; Doraiswamy et al., 2020; Torous & Blease, 2024).”

“This vision aligns with broader healthcare innovations that employ machine learning to refine prediction accuracy and personalise care pathways (Al-Worafi et al., 2023; Pham et al., 2022).”

“Users are more likely to adopt digital systems when they believe that ethical standards, data privacy, and accountability are well established (Balaskas et al., 2025; Choudhury et al., 2025).”

“Ease of interaction has also been shown to reduce anxiety around technology use and encourage sustained engagement, especially when conversational agents maintain human-like empathy and clarity (Babu & Joseph, 2024).”

“Such adaptability aligns with earlier findings that natural-language models, such as ChatGPT, can generate empathetic and contextually sensitive responses that enhance engagement and emotional support (Elyoseph et al., 2023; Howcroft et al., 2025).”

“In mental health contexts, simple, clear, and responsive conversational systems help create a sense of comfort and psychological safety for both clinicians and patients (Heston, 2023).”

“This study employed thematic analysis to systematically and credibly capture and interpret expert perspectives (Naisola-Ruiter, 2022).”

“Previous research has similarly underscored that conversational AI can deliver empa-thetic, language-based interactions that facilitate early symptom identification and emotional regulation (Balcombe, 2023; Ettman & Galea, 2023).”

“A two-round Delphi technique was employed to facilitate iterative data collection and analysis (Tee et al., 2022).”

“By using natural language processing and machine learning, these tools can improve emotional regulation and user engagement in mental health self-care (Elyoseph et al., 2023; Liu et al., 2023).”

“The ranking procedure required experts to assign unique numerical values to each item, ensuring clarity, eliminating duplicate rankings, and enhancing statistical validity (Parker et al., 2021).”

“Over the years, TAM has been expanded to include additional dimensions such as trust, risk perception, and ethical awareness, particularly in fields involving sensitive human interaction such as healthcare (Ko & Leem, 2021; Ratta et al., 2025).”

“…rather than as replacements, especially for individuals with severe or complex mental health conditions and for those living in areas with limited access to qualified professionals (Blease & Torous, 2023; Essien & Asamoah, 2020; Miner et al., 2020; Patel et al., 2007; Torous & Blease, 2024).”

“Users are more likely to adopt digital systems when they believe that ethical standards, data privacy, and accountability are well established (Balaskas et al., 2025; Choudhury et al., 2025).”

“GAI, with its ability to conduct lifelike conversations, introduces a new form of self-guided online intervention for individuals experiencing symptoms of depression and anxiety (Pandya et al., 2023).”

“These findings align with earlier work, which has shown that AI-assisted responses can enhance both accuracy and empathy in patient interactions (Tate et al., 2023).”

“Future policies should therefore prioritise equitable usability, reliability, and ethical compliance to fulfil the democratising promise of GAI (Allen & Woodnutt, 2023).”

“However, occasional lapses in medical precision were noted (Kharko et al., 2024).”

“GAI applications have been developed to provide counselling and emotional support, assist in early detection of symptoms, and promote psychoe-ducation (Liu et al., 2023; Srivastava & Srivastava, 2023).”

“Empirical studies in healthcare and AI adoption consistently find that trust mediates between user intention and acceptance, while perceived risk negatively influences acceptance (Goh et al., 2024; Kauttonen et al., 2025).”

“Evidence suggests that when clinicians and patients perceive AI systems as helpful in improving therapeutic efficiency and the quality of emotional care, their intention to adopt and continue using these tools increases significantly (Békés et al., 2025).”

“Recent studies emphasise that transparency, fairness, and accountability are essential for trustworthy AI systems that protect patient autonomy and privacy while maintaining clinician–patient trust (Cheong, 2024; Ott & Dabrock, 2022).”

“In mental health care, where confidentiality and empathy are vital, building transparent and ethically governed AI systems is a prerequisite for clinical acceptance and long-term integration into professional practice (Saeidnia et al., 2024; Svensson et al., 2025).”

“In the context of mental health, positive emotional experiences when interacting with AI-assisted tools can strengthen engagement and trust (Peng et al., 2025).”

“Within the field of mental health, GAI applications such as ChatGPT are receiving increasing attention for their ability to provide personalised support, facilitate therapeutic interactions, and assist in mental health assessments (Aditama et al., 2023).”

“These insights align with frameworks in learning health systems and precision public health that leverage data-driven approaches to improve intervention design (Elyoseph et al., 2023; Friedman et al., 2015; Xian, 2024).”

“Studies have argued that mental health professionals must actively shape how AI is integrated into psychiatric care to ensure that its deployment aligns with ethical, clinical, and empathic standards (Moran, 2024).”

“In addition, limited attention has been given to caregiver-centered questions, multimorbidity, polypharmacy, cognitive impairment, self-neglect, and the distinction between depressive symptoms and acute medical deterioration (32).”