Empathic AI for Patient-Centered Cancer Care: A Scoping Review of Patient Navigation, Support, and Clinical Practice

May combine two claims.

In oncology, AI models are being developed to standardize and improve the consistency and quality of care and to support clinicians and patients in navigating complex diagnostic and treatment pathways.

“In oncology, AI models are being developed to standardize and improve the consistency and quality of care, offering new ways to support both clinicians and patients in navigating complex diagnostic and treatment pathways [5].”

• cites [5] Kolla, Uses and limitations of artificial intelligence for oncology, Cancer sourcePartly supported

  The Introduction says AI can improve cancer screening, diagnosis, and treatment decisions, while the Chatbots section describes potential support for patients and clinicians. The paper does not establish that AI models are being developed to standardize care or improve its consistency; its Discussion instead calls for standardized data collection and reliable algorithms. Page numbers are unavailable in our copy. (LLM, conf=0.83)

  Experimental — accuracy not yet measured

May combine two claims.

AI technologies provide timely, accurate, and accessible information that can empower patients and caregivers, supporting engagement, comprehension, and autonomy throughout the cancer care continuum.

“These technologies provide timely, accurate, and accessible information that can empower patients and caregivers, supporting engagement, comprehension, and autonomy throughout the cancer care continuum [6-8].”

• cites [6] Cè, Exploring the role of ChatGPT in oncology: providing information and support for cancer patients, BioMedInformatics sourcePartly supported

  Sections 3.1.2 and 3.1.3 describe ChatGPT as making information easier and faster to obtain and understand, including for patients and caregivers, with potential to improve participation and shared decision-making. The Conclusions say it can offer accurate, accessible support throughout patients’ cancer journey, but also warn that its information can be inaccurate or unreliable. The claim overstates accuracy as an unqualified property of AI technologies generally. (LLM, conf=0.82)

  Experimental — accuracy not yet measured
• cites [7] Li, The impact of Chat Generative Pre-trained Transformer (ChatGPT) on oncology: application, expectations, and future prospects, Cureus sourcePartly supported

  The Review’s “ChatGPT in cancer patient education” section describes the potential for accessible explanations that improve patients’ comprehension, and its discussion of treatment decisions says such information can support engagement. It does not establish the same benefits for caregivers or throughout the cancer care continuum. Its “Rigorous Fact-Checking” section also cautions that responses may be outdated or erroneous, so timely, accurate information cannot be assumed. (LLM, conf=0.86)

  Experimental — accuracy not yet measured
• cites [8] Goel, Role of AI in empowering and redefining the oncology care landscape: perspective from a developing nation, Front Digit Health sourcePartly supported

  The ‘AI in telemedicine/remote monitoring’ section (page unavailable) describes apps that give cancer patients individualized treatment information, advice, reminders, and access to support; the oncology-care table also describes improved access to services. Our copy does not establish that this information is consistently accurate or that these tools improve caregivers’ comprehension, engagement, or autonomy throughout the cancer care continuum. (LLM, conf=0.84)

  Experimental — accuracy not yet measured

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

AI has become increasingly embedded in medical practice over the past decade, enhancing care capabilities from personalized health libraries and patient education to disease screening, diagnostics, and treatment planning.

“Artificial intelligence (AI) has become increasingly embedded in medical practice over the past decade, enhancing care capabilities from the development of personalized health libraries and innovative patient education strategies to disease screening, diagnostics, and treatment planning [1-4].”

• cites [1] Khosravi, Artificial intelligence and decision-making in healthcare: a thematic analysis of a systematic review of reviews, Health Serv Res Manag Epidemiol sourcePartly supported

  The Introduction describes AI applications in diagnosis and treatment planning, and the Discussion > Shared Decision-Making describes personalized information and patient education (page numbers unavailable). The copy does not establish the specific development of personalized health libraries or that AI became increasingly embedded in medical practice over the past decade. (LLM, conf=0.85)

  Experimental — accuracy not yet measured
• cites [2] Shaban-Nejad, Health intelligence: how artificial intelligence transforms population and personalized health, NPJ Digit Med sourcePartly supported

  The Abstract (page unavailable) describes AI applications in patient education, medical imaging, and decision support; Personalized health intelligence describes tailoring therapy plans and connecting patients with information. The copy does not establish a trend specifically over the past decade or describe the development of personalized health libraries. It discusses diagnosis and surveillance, but does not clearly identify disease screening as an application. (LLM, conf=0.85)

  Experimental — accuracy not yet measured
• cites [3] Ammar, Using a personal health library-enabled mHealth recommender system for self-management of diabetes among underserved populations: use case for knowledge graphs and linked data, JMIR Form Res sourcePartly supported

  The Introduction (Overview; page unavailable) describes personal health libraries that can generate intelligent personalized interventions, and the Discussion (page unavailable) proposes AI-based personalized resource suggestions. The Review of Relevant Literature also mentions a tool for cancer staging and identifying treatment options. The paper does not establish a decade-long increase in AI use across medical practice or support the full range of screening and diagnostic capabilities attributed to the citation group. (LLM, conf=0.83)

  Experimental — accuracy not yet measured
• cites [4] Shaban-Nejad A Michalowski M Bianco S . Multimodal AI in Healthcare: A Paradigm Shift in Health Intelligence. Springer; 2023. [doi: 10.1007/978-3-031-14771-5] sourceNot checked — no full text of the source

May combine two claims.

By integrating psychosocial considerations into clinical decision-making, AI holds promise for improving treatment efficacy and enhancing overall quality of life.

“By integrating psychosocial considerations into clinical decision-making, AI holds promise not only for improving treatment efficacy but also for enhancing overall quality of life [10].”

• cites [10] Auf, The use of AI in mental health services to support decision-making: scoping review, J Med Internet Res sourcePartly supported

  The Introduction (Background; page unavailable) describes AI’s potential for personalized clinical decision-making, and the Results (Evaluation of Implementation and Use; page unavailable) describe its potential to support treatment selection and communication. Those passages do not link the integration of psychosocial considerations to improved treatment efficacy or overall quality of life; searches for those terms found no such account in our copy. (LLM, conf=0.82)

  Experimental — accuracy not yet measured

May combine two claims.

Greer's Vivibot chatbot leveraged a positive psychology framework to support emotional regulation in young adult cancer survivors, resulting in reduced anxiety and high user satisfaction.

“Greer’s Vivibot chatbot leveraged a positive psychology framework to support emotional regulation in young adult cancer survivors, resulting in reduced anxiety and high user satisfaction [14].”

• cites [14] Greer, Use of the chatbot “Vivibot” to deliver positive psychology skills and promote well-being among young people after cancer treatment: randomized controlled feasibility trial, JMIR Mhealth Uhealth sourcePartly supported

  The Methods (Intervention Conditions) describe Vivibot’s positive-psychology skills, and the Results report helpfulness ratings and an average reduction in anxiety in the intervention group. However, the between-group anxiety finding was only a nonsignificant trend (P=.09), so “resulting in reduced anxiety” implies a firmer causal finding than this feasibility trial establishes. (LLM, conf=0.88)

  Experimental — accuracy not yet measured

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

AI has become a critical adjunct for medical image analytics, with computer models achieving state-of-the-art tumor detection, segmentation, extraction, and classification across a wide range of cancer types.

“For example, AI has rapidly become a critical adjunct for medical image analytics, demonstrating that computer models can achieve state-of-the-art tumor detection, segmentation, extraction, and classification across a wide range of cancer types [9].”

• cites [9] Debelee, Deep learning in selected cancers’ image analysis—a survey, J Imaging sourceSupported

May combine two claims.

Recent advancements suggest that AI systems may approximate emotional attunement, recognize distress, and respond in ways perceived as empathic and compassionate.

“Yet recent advancements suggest that AI systems may approximate emotional attunement, recognize distress, and respond in ways perceived as empathic and compassionate [13].”

• cites [13] Rubin, Considering the role of human empathy in AI-driven therapy, JMIR Ment Health sourceSupported

May combine two claims.

Chaix et al and Siglen et al reported that chatbot-delivered education improved adherence to treatment regimens and facilitated greater engagement in shared decision-making.

“Similarly, Chaix et al [17] and Siglen et al [18] reported that chatbot-delivered education improved adherence to treatment regimens and facilitated greater engagement in shared decision-making.”

• cites [17] Chaix, When chatbots meet patients: one-year prospective study of conversations between patients with breast cancer and a chatbot, JMIR Cancer sourcePartly supported

  The Discussion (Conclusions; page unavailable) says Vik improved medication adherence through reminders and educational content, and the Results report improved compliance among patients using the reminder feature. The paper does not report greater engagement in shared decision-making, so that part of the claim is unsupported. (LLM, conf=0.88)

  Experimental — accuracy not yet measured
• cites [18] Siglen, Ask Rosa—the making of a digital genetic conversation tool, a chatbot, about hereditary breast and ovarian cancer, Patient Educ Couns sourceCould not verify

  The Introduction and Conclusions describe Ask Rosa as a source of genetic information and suggest it may help patients prepare for counseling. Our copy does not report improved treatment adherence or greater engagement in shared decision-making; searches for both outcomes found no supporting passage. (LLM, conf=0.76)

  Experimental — accuracy not yet measured

May combine two claims.

Progress on interpersonal-type communications is being made, despite challenges noted on the algorithmic development side.

“progress on interpersonal-type communications is being made, despite challenges noted on the algorithmic development side [45].”

• cites [45] Chow, Ethical considerations in human-centered AI: advancing oncology chatbots through large language models, JMIR Bioinform Biotechnol sourceSupported

May combine two claims.

ChatGPT has been used to assist in the interpretation of personalized breast cancer screening studies with 80% accuracy while reducing clinician workload by 70%.

“For example, ChatGPT has been used to assist in the interpretation of personalized breast cancer screening studies with 80% accuracy while reducing clinician workload by 70% [38].”

• cites [38] Mundinger, Artificial intelligence in senology—where do we stand and what are the future horizons?, Eur J Breast Health sourcePartly supported

  The section ‘Have These Possibilities Already Touched the Field of Breast Healthcare?’ (page not supplied) reports that ChatGPT recommendations, including radiology-screening recommendations, were at least 80% in line with human judgement. It does not report 80% accuracy in interpreting personalized breast cancer screening studies. The ‘AI in Breast Cancer Screening’ section reports a 44.3% reduction in screen-reading workload with AI-assisted mammography, not a 70% reduction attributable to ChatGPT. (LLM, conf=0.85)

  Experimental — accuracy not yet measured

May combine two claims.

Recent work explored hybrid and simulation-based approaches, including live-chat models emphasizing digital clinical empathy and virtual patient systems designed to train clinicians in empathic and end-of-life communication.

“Recent work has also explored hybrid and simulation-based approaches, including live-chat models emphasizing digital clinical empathy [34] and virtual patient systems designed to train clinicians in empathic and end-of-life communication [3536].”

• cites [34] Luetke Lanfer, Digital clinical empathy in a live chat: multiple findings from a formative qualitative study and usability tests, BMC Health Serv Res sourceSupported
• cites [3536]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

Socially intelligent agents can address nonmedical needs in pediatric populations by reducing feelings of isolation, supporting coping, and facilitating participation in daily activities.

“Mårell-Olsson et al [44] demonstrated that socially intelligent agents can address nonmedical needs in pediatric populations by reducing feelings of isolation, supporting coping, and facilitating participation in daily activities.”

• cites [44] Mårell-Olsson, Opportunities and challenges of using socially intelligent agents: increasing interaction and school participation for children suffering from a long-term illness, IJILT sourcePartly supported

  The Introduction and Discussion and conclusions identify isolation and school participation as nonmedical concerns for children with long-term illness. The paper also proposes SIAs as possible support for interaction and healthy coping. But its evidence comes from expert interviews, not a demonstration that SIAs reduced isolation or improved coping and daily participation in children; the Discussion says the small study cannot support generalizations. (LLM, conf=0.87)

  Experimental — accuracy not yet measured

May combine two claims.

One social media-based study found that nonhuman responses were rated higher on every measure by expert health care evaluators, including empathy and quality, and greatly preferred to the clinician responses.

“one social media–based study found that nonhuman responses were rated higher on every measure by expert health care evaluators, including empathy and quality, and greatly preferred to the clinician responses [32].”

• cites [32] Ayers, Comparing physician and artificial intelligence chatbot responses to patient questions posted to a public social media forum, JAMA Intern Med sourceSupported

May combine two claims.

AI personalized conversations and treatment recommendations help patients make informed decisions and improve treatment adherence.

“This helps patients make informed decisions and improves treatment adherence [57].”

• cites [57] Klotz, ChatGPT vs. surgeons on pancreatic cancer queries: accuracy & empathy evaluated by patients and experts, HPB (Oxford) sourcePartly supported

  The Discussion reports that ChatGPT’s answers to general pancreatic-cancer surgery questions were understandable and showed promise for patient education. It does not establish that personalized conversations or treatment recommendations helped patients make informed decisions, and it did not assess treatment adherence. (LLM, conf=0.80)

  Experimental — accuracy not yet measured

Probably not a substantial claim: may not pass S3 (load-bearing).

Sentiment-aware conversational agents are designed to detect and respond to emotional distress.

“sentiment-aware conversational agents designed to detect and respond to emotional distress [49].”

• cites [49] Hamza-Cherif, Improving healthcare communication: AI-driven emotion classification in imbalanced patient text data with explainable models, Adv Technol Innov sourcePartly supported

  The Introduction describes AI models that detect patients’ feelings from text, and Additional information discusses identifying negative thoughts. Neither passage describes a conversational agent or says the system responds to emotional distress; the supported part is detection, not response. (LLM, conf=0.87)

  Experimental — accuracy not yet measured

May combine two claims.

Digital clinical empathy requires intentional strategies (timely acknowledgment, authenticity, appropriate boundary setting, contextual responsiveness) and may be best sustained through hybrid models combining AI-enabled support with human oversight.

“Studies measuring cancer-related live chat environments and simulation-based systems highlight that digital clinical empathy requires intentional strategies (ie, timely acknowledgment, authenticity, appropriate boundary setting, and contextual responsiveness) and may be best sustained through hybrid models that combine AI-enabled support with human oversight [34-36].”

• cites [34] Luetke Lanfer, Digital clinical empathy in a live chat: multiple findings from a formative qualitative study and usability tests, BMC Health Serv Res sourceSupported
• cites [35] Kane, Managing emotional dialogue for a virtual cancer patient: a schema-guided approach, IEEE Trans Affective Comput sourceNot checked — no full text of the source
• cites [36] Ali, Novel computational linguistic measures, dialogue system and the development of SOPHIE: standardized online patient for healthcare interaction education, IEEE Trans Affective Comput sourcePartly supported

  Split into 2 atoms — 0 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 1 unverifiable. Expand the claim to see per-atom verdicts.

  Experimental — accuracy not yet measured

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

LLMs have the potential to provide real-time, individually customized, and up-to-date medical information, with concomitant risks and benefits, to allow for informed, complex, and empathic decision-making among caregivers for oncology patients.

“Of particular interest is the potential of LLMs as providers of real-time, individually customized, and up-to-date medical information, with their concomitant risks and benefits, to allow for informed, complex, and empathic decision-making among caregivers for oncology patients [30].”

• cites [30] Sezgin, Can large language models aid caregivers of pediatric cancer patients in information seeking? A cross-sectional investigation, Cancer Med sourcePartly supported

  Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  Experimental — accuracy not yet measured

May combine two claims.

Statistically significant improvements were found in the psychological measures of mood and emotional well-being, with overall reductions in stress levels demonstrated due to the safe and soothing environment simulated by the intervention.

“Statistically significant improvements were found in the psychological measures of mood and emotional well-being, with overall reductions in stress levels demonstrated due to the safe and soothing environment simulated by the intervention.”

• cites [51] O’Gara, SafeSpace: what is the feasibility and acceptability of a codesigned virtual reality intervention, incorporating compassionate mind training, to support people undergoing cancer treatment in a clinical setting?, BMJ Open sourcePartly supported

  The Results/findings > Multivariate analyses section (page unavailable) reports statistically significant improvements in mood after the first session, mental well-being, and stress from baseline to after session 3; mood improvement after the second session was not statistically significant. The Discussion (page unavailable) says participants valued the simulated ‘safe space,’ but proposes a future trial to determine whether VR, compassionate mind training, or both produced the effects. The claim’s attribution of the stress reduction specifically to the safe and soothing environment is therefore not established. (LLM, conf=0.88)

  Experimental — accuracy not yet measured

May combine two claims.

Empathic AI interventions improved mood, stress reduction, and support in palliative and end-of-life contexts.

“Improved mood, stress reduction, and support in palliative and end-of-life contexts”

May combine two claims.

Empathic AI applications for enhanced diagnostics and treatment planning include AI-assisted screening, documentation, and decision-making.

“AIc-assisted screening, documentation, and decision-making”

May combine two claims.

Concerns were identified regarding overreliance on AI systems, cultural insensitivity, and patient privacy.

“However, concerns were identified regarding overreliance on AI systems, cultural insensitivity, and patient privacy.”

May combine two claims.

Chatbots and AI cofacilitators were most often used to simplify complex medical concepts, offer condition-specific coping strategies, and reduce distress through emotionally calibrated messaging.

“Chatbots and AI cofacilitators were most often used to simplify complex medical concepts, offer condition-specific coping strategies, and reduce distress through emotionally calibrated messaging.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The initial electronic database search yielded 78 results.

“The initial electronic database search yielded 78 results”

May combine two claims.

Common applications of empathic AI include personalized education for patients and providers, support for diagnostic and therapeutic decisions, and tools designed to recognize and respond to patient distress.

“Common applications include personalized education for patients and providers, support for diagnostic and therapeutic decisions, and tools designed to recognize and respond to patient distress.”

May combine two claims.

Empathic AI applications for tailored communication and education include real-time summary tools and simulation-based training.

“Real-time summary tools and simulation-based training”

Probably not a substantial claim: may not pass S3 (load-bearing).

A snowballing technique identified 15 additional studies.

“This process led to the identification of 15 additional studies”

Probably not a substantial claim: may not pass S3 (load-bearing).

The included literature spans from June 2018 to January 2025.

“The included literature spans from June 2018 to January 2025”

May combine two claims.

In included studies, empathic AI applications for enhanced diagnostics and treatment planning improved adherence, patient satisfaction, and provider-patient trust.

“Improved adherence, patient satisfaction, and provider-patient trust”

May combine two claims.

Several studies noted improved patient engagement and reduced clinician burden.

“Several studies noted improved patient engagement and reduced clinician burden.”

May combine two claims.

In included studies, tailored communication and education applications of empathic AI were reported to improve patient comprehension, reduce distress, and increase engagement in care.

“Improved patient comprehension, reduced distress, and greater engagement in care”

May combine two claims.

Empathic AI interventions for emotional and psychosocial support reduced anxiety, isolation, and emotional distress in patients.

“Reduced anxiety, isolation, and emotional distress in patients”

Probably not a substantial claim: may not pass S3 (load-bearing).

Following abstract and full-text screening, 64 studies were eliminated because they did not meet the inclusion criteria.

“Following the abstract and full-text screening, 64 studies were eliminated because they did not meet the inclusion criteria.”

May combine two claims.

In the included studies, VRd interventions and NLP tools were used for end-of-life discussions.

“VRd interventions and NLP tools for end-of-life discussions”

May combine two claims.

A total of 44 studies were included and analyzed thematically across 3 core clinical domains: tailored communication and education, diagnostics and care plan optimization, and emotional and psychosocial support.

“A total of 44 studies were included and analyzed thematically across 3 core clinical domains: tailored communication and education, diagnostics and care plan optimization, and emotional and psychosocial support.”

• Also stated in: § Overview · plain

  “a total of 44 studies met the inclusion criteria for review.”

May combine two claims.

In total, 70% (31/44) of the included studies addressed the use of empathic AI in the clinical space, with approximately half examining both diagnostics and treatment planning.

“In total, 70% (31/44) of the included studies addressed the use of empathic AI in the clinical space, with approximately half examining both diagnostics and treatment planning.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Several studies raised concerns about how increasing reliance on AI may shift the balance between algorithmic input and clinical judgment, prompting debate about the potential erosion or recalibration of clinical intuition.

“Several studies raised concerns about how increasing reliance on AI may shift the balance between algorithmic input and clinical judgment, prompting debate about the potential erosion or recalibration of clinical intuition.”

May combine two claims.

AI-generated guidance influences treatment decision-making in advanced or complex cancer contexts, with variability in tone, content, and recommendations that may shape patient choices.

“Others emphasize the influence of AI-generated guidance on treatment decision-making in advanced or complex cancer contexts, including variability in tone, content, and recommendations that may shape patient choices [2252].”

• cites [2252]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

Other studies explored the use of empathic AI for real-time feedback and simulation-based training, with AI tools positioned as companions in ethical and emotional reasoning.

“Other studies explored the use of empathic AI for real-time feedback and simulation-based training, with AI tools positioned not just as repositories of information but as companions in ethical and emotional reasoning [2627].”

• cites [2627]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

NLP models have been used to analyze patients' decision-making behaviors and emotions, allowing clinicians to understand patients better and tailor their recommendations accordingly.

“Furthermore, NLP models have been used to analyze patients’ decision-making behaviors and emotions [3341], allowing clinicians to understand patients better and tailor their recommendations accordingly.”

• cites [3341]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

AI enhances patient-provider communication and real-time emotional support, leading to optimized care plans.

“Studies included in this theme demonstrate that AI enhances patient-provider communication and real-time emotional support, leading to optimized care plans.”

Probably not a substantial claim: may not pass S3 (load-bearing).

One study highlighted that empathic AI may be particularly important in high-stress oncology settings for reducing clinician burnout and compassion fatigue.

“As 1 study highlighted, this may be particularly important in high-stress oncology settings for reducing clinician burnout and compassion fatigue [29].”

• cites [29] Koranteng, Compassionate machines: the ethics of “artificial empathy” in cancer care, JAMA Oncol sourceNot checked — no full text of the source

May combine two claims.

Providers received AI-generated conversation summaries or context-sensitive prompts, which improved their ability to convey difficult information with empathy and precision.

“Providers received AI-generated conversation summaries or context-sensitive prompts, which improved their ability to convey difficult information with empathy and precision.”

May combine two claims.

Several studies focused on the ways empathic AI can provide emotional comfort to patients, offering them a sense of companionship and reducing feelings of isolation.

“Several studies focused on the ways empathic AI can provide emotional comfort to patients, offering them a sense of companionship and reducing feelings of isolation.”

Classified as probably not a substantial claim. May combine two claims.

Although AI may reduce cognitive load and offer scalable training solutions, its ability to cultivate true empathic capacity remains contested.

“Although AI may reduce cognitive load and offer scalable training solutions, its ability to cultivate true empathic capacity remains contested”

May combine two claims.

Studies in this theme revealed that AI systems are increasingly integrated into clinical decision-making, optimizing care goals and improving patient outcomes.

“Studies in this theme revealed how AI systems are increasingly integrated into clinical decision-making, optimizing care goals and improving patient outcomes.”

May combine two claims.

These tools were positioned as cognitive extenders, offering context-specific recommendations on treatment protocols, synthesizing complex medical data, and facilitating more empathic responses during patient interactions.

“These tools were positioned as cognitive extenders, offering context-specific recommendations on treatment protocols, synthesizing complex medical data, and facilitating more empathic responses during patient interactions.”

May combine two claims.

AI applications offer tailored resources, emotional support, and coping strategies, helping caregivers manage their own well-being while supporting their loved ones through cancer care.

“AI applications were found to offer tailored resources, emotional support, and coping strategies, helping caregivers manage their own well-being while supporting their loved ones through cancer care.”

• AssumptionRests on an assumption: Caregivers often experience high levels of stress and burnout.

May combine two claims.

These algorithms have supported individualized radiation therapy dosage planning, enhancing diagnostic efficiency and precision in a patient-centered manner.

“Similarly, these algorithms have supported individualized radiation therapy dosage planning [39], enhancing diagnostic efficiency and precision in a patient-centered manner.”

• cites [39] Turner, Human-artificial intelligence symbiotic reporting for theranostic cancer care, Cancer Biother Radiopharm sourceNot checked — no full text of the source

May combine two claims.

Empathic AI-generated responses may vary in tone, emotional sensitivity, or guidance depending on prompt structure, clinical framing, or patient demographics, potentially shaping patient perceptions, fostering overtrust or skepticism, and influencing clinical decision-making.

“Empathic AI-generated responses may vary in tone, emotional sensitivity, or guidance depending on prompt structure, clinical framing, or patient demographics, potentially shaping patient perceptions, fostering over trust or skepticism, and influencing clinical decision-making.”

May combine two claims.

Emerging work in sentiment modeling and empathic conversational agent development suggests early methodological movement toward quantifying affective signals, although approaches remain heterogeneous and not yet clinically standardized.

“At the same time, emerging work in sentiment modeling and empathic conversational agent development suggests early methodological movement toward quantifying affective signals, although approaches remain heterogeneous and not yet clinically standardized [495354].”

• cites [495354]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

This review highlights early promise and high user receptivity.

“While this review highlights early promise and high user receptivity, clinical integration must be guided by rigorous validation, ethical oversight, and a deep understanding of real-world workflows and health system readiness.”

May combine two claims.

Generative AI responses to cancer-related decision-making scenarios may vary by patient demographics and clinical framing, with inconsistencies in tone, content, and guidance that could differentially influence patient perceptions and treatment decisions.

“Evidence suggests that generative AI responses to cancer-related decision-making scenarios may vary by patient demographics and clinical framing, with inconsistencies in tone, content, and guidance that could differentially influence patient perceptions and treatment decisions, underscoring the need for careful validation and health equity–focused oversight [2224].”

• cites [2224]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

Some studies used patient satisfaction or emotional resonance as proxies, while others relied on subjective expert evaluation.

“Some studies used patient satisfaction or emotional resonance as proxies, while others relied on subjective expert evaluation.”

Classified as probably not a substantial claim.

A consistent theme throughout the examined literature is the redefinition of what it means to deliver holistic care.

“A consistent theme throughout the examined literature is the redefinition of what it means to deliver holistic care.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

AI technologies have ushered in a new era of precision health and medicine, specifically impacting oncology, informing care strategies from recurrence prediction and surgical planning to emotional assessment.

“These technologies have ushered in a new era of precision health and medicine [11], specifically impacting the field of oncology, informing care strategies that range from recurrence prediction and surgical planning to emotional assessment.”

• cites [11] Shaban-Nejad A Michalowski M . Precision Health and Medicine: A Digital Revolution in Healthcare. Springer; 2020. [doi: 10.1007/978-3-030-24409-5] sourceNot checked — no full text of the source

May combine two claims.

Empathic AI demonstrates the potential to improve cancer care by enhancing patient education, clinical decision-making, and emotional support.

“Empathic AI demonstrates the potential to improve cancer care by enhancing patient education, clinical decision-making, and emotional support.”

Probably not a substantial claim: may not pass S3 (load-bearing).

A third reviewer verified the full set of included and excluded studies, cross-referencing decisions with the original criteria and extraction tables.

“A third reviewer (BW) verified the full set of included and excluded studies, cross-referencing decisions with the original criteria and extraction tables.”

Probably not a substantial claim: may not pass S3 (load-bearing).

A systematic search of literature was conducted in accordance with PRISMA 2020 guidelines.

“A systematic search of literature was conducted in accordance with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 guidelines.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Study quality indicators (peer-review status, study design, and sample size) were collected during data extraction to contextualize findings and mitigate potential sources of bias.

“Study quality indicators (ie, peer-review status, study design, and sample size) were collected during data extraction to contextualize findings and mitigate potential sources of bias.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

This progress may help address systemic gaps in access to emotional support and culturally responsive communication in cancer care.

“This progress may help address systemic gaps in access to emotional support and culturally responsive communication, which have been long-standing public health challenges in cancer care.”

• AssumptionRests on an assumption: There are systemic gaps in access to emotional support and culturally responsive communication in cancer care that are long-standing public health challenges.

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

Empathic AI represents a promising advancement in patient-centered oncology, integrating emotional intelligence into clinical care.

“Empathic AI represents a promising advancement in patient-centered oncology, integrating emotional intelligence into clinical care.”

May combine two claims.

While AI integration into the clinical realm is in its infancy, it has a definite role in enhancing patient care delivery and optimization.

“While AI integration into the clinical realm is in its infancy, it has a definite role in enhancing patient care delivery and optimization.”

May combine two claims.

AI-based therapy group cofacilitator allows physicians to tailor patient-to-provider interactions, resulting in more patient-centered care.

“which allows physicians to tailor their patient-to-provider interactions accordingly, resulting in more patient-centered care.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Scalability and safety remain central to the integration of AI-mediated education in oncology as these tools move closer to frontline deployment.

“Scalability and safety remain central to the integration of AI-mediated education in oncology as these tools move closer to frontline deployment.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

Together, these functions of AI platforms create not only a precise and tailored patient experience but also one that is more patient-specific.

“Together, these functions of AI platforms create not only a precise and tailored patient experience but also one that is more patient-specific.”

Classified as probably not a substantial claim.

In oncology, missteps from biased AI could carry particularly high stakes because communication must be attuned to cultural norms, family dynamics, and existential concerns.

“In oncology, where communication must be attuned to cultural norms, family dynamics, and existential concerns, these missteps could carry particularly high stakes.”

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

Overreliance on AI-generated outputs risks erosion of clinicians' empathic instincts, raising concerns about long-term implications for professional development and patient trust.

“The risk of overreliance on AI-generated outputs and the potential erosion of clinicians’ empathic instincts raise critical concerns about long-term implications for professional development and patient trust.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Some models and interventions included in this review may already be outdated or replaced by more sophisticated systems.

“Given the pace of advancement in the field, some of the models and interventions included in this review may already be outdated or replaced by more sophisticated systems.”

• AssumptionRests on an assumption: AI technologies are rapidly evolving.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Historically underserved populations or individuals with low health literacy may benefit most from scalable, emotionally attuned technologies.

“This is especially significant for historically underserved populations or individuals with low health literacy, who may benefit most from scalable, emotionally attuned technologies.”

Probably not a substantial claim: may not pass S2 (checkable).

Successful integration of empathic AI into clinical workflows depends on careful design, robust training data, and a commitment to maintaining the human essence of medicine.

“Effective implementation will depend on careful design, robust training data, and a commitment to maintaining the human essence of medicine.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Poorly integrated AI tools may increase cognitive load, disrupt clinical workflows, or generate alert fatigue, undermining adoption and trust.

“Poorly integrated AI tools may increase cognitive load, disrupt clinical workflows, or generate alert fatigue, undermining adoption and trust.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The heterogeneity of included studies reflects the early and exploratory nature of the field.

“although this heterogeneity reflects the early and exploratory nature of the field.”

May combine two claims.

Empathic AI has the potential to enhance not only the efficacy and accuracy of care but also its psychosocial dimensions.

“empathic AI has the potential to enhance not only the efficacy and accuracy of care but also its psychosocial dimensions.”

May combine two claims.

AI expands access to personalized education and emotional support, democratizing information and empowering users to take an active role in their treatment.

“For patients, AI expands access to personalized education and emotional support, democratizing information and empowering users to take an active role in their treatment.”

May combine two claims.

AI-generated conversation summaries, real-time feedback, and emotion-sensitive prompts can enhance provider-patient dialogues and reduce burnout by offloading cognitive and emotional labor.

“AI-generated conversation summaries, real-time feedback, and emotion-sensitive prompts can enhance provider-patient dialogues and reduce burnout by offloading cognitive and emotional labor, such as the analysis of patient decision-making and emotions provided by the PRIME-2 (Patient Reported Information Multidimensional Exploration version 2) framework [46].”

• cites [46] Ranasinghe, Robotic-assisted vs. open radical prostatectomy: a machine learning framework for intelligent analysis of patient-reported outcomes from online cancer support groups, Urol Oncol sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

LLMs can be embedded into electronic health record systems to generate on-demand phrasing suggestions for emotionally sensitive conversations, such as treatment failure or end-of-life planning.

“LLMs can be embedded into electronic health record systems to generate on-demand phrasing suggestions for emotionally sensitive conversations, such as treatment failure or end-of-life planning.”

Probably not a substantial claim: may not pass S2 (checkable).

Overreliance on AI in emotionally sensitive encounters could lead to depersonalization or erosion of therapeutic rapport, particularly if patients perceive machine-generated responses as inauthentic or misaligned with their values.

“Overreliance on AI in emotionally sensitive encounters could lead to depersonalization or erosion of therapeutic rapport between patient and provider, particularly if patients perceive machine-generated responses as inauthentic or misaligned with their values.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Generative AI systems may produce fluent, confident, but clinically incorrect or incomplete responses, which can be particularly harmful in oncology decision-making and end-of-life contexts.

“Generative AI systems may produce fluent, confident, but clinically incorrect or incomplete responses, which can be particularly harmful in oncology decision-making and end-of-life contexts.”

May combine two claims.

This review demonstrates that empathic AI is emerging as a powerful and transformative force within oncology, with broad clinical applications from education to diagnostics and treatment planning to emotional support.

“This review demonstrates that empathic AI is emerging as a powerful and transformative force within oncology, with broad clinical applications from education to diagnostics and treatment planning to emotional support.”

Probably not a substantial claim: may not pass S4 (based).

Safety is further complicated by the risk of misinformation or emotionally manipulative outputs.

“Safety is further complicated by the risk of misinformation or emotionally manipulative outputs.”

May combine two claims.

From a provider perspective, AI serves as both a diagnostic and treatment decision support system and a partner in empathic communication.

“From a provider perspective, AI serves as both a diagnostic and treatment decision support system and a partner in empathic communication.”

Classified as probably not a substantial claim.

Current tools may not yet meet the usability, safety, or integration standards required for large-scale adoption.

“Current tools may not yet meet the usability, safety, or integration standards required for large-scale adoption, underscoring the need for ongoing evaluation in real-world settings.”

Probably not a substantial claim: may not pass S4 (based).

Empathic AI systems often process highly sensitive emotional, psychosocial, and behavioral data, increasing risks related to privacy, consent, and secondary data use.

“Empathic AI systems often process highly sensitive emotional, psychosocial, and behavioral data, increasing risks related to privacy, consent, and secondary data use.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

AI's ability to contextualize medical information, detect distress, and respond with calibrated empathy suggests that these systems can complement and bolster human care, particularly in emotionally charged settings such as diagnosis disclosure, palliative care, and end-of-life decision-making.

“AI’s ability to contextualize medical information, detect distress, and respond with calibrated empathy suggests that these systems can complement and bolster human care, particularly in emotionally charged settings such as diagnosis disclosure, palliative care, and end-of-life decision-making.”

Classified as probably not a substantial claim.

Sociocultural dynamics shape both the reception and use of empathic AI.

“Moreover, sociocultural dynamics shape both the reception and use of empathic AI.”

• AssumptionRests on an assumption: Concepts of empathy, emotional expression, and caregiving may vary considerably across cultures; what is perceived as compassionate in one context may be viewed as intrusive or inappropriate in another.

May combine two claims.

Empathic AI systems show promise in extending the reach of clinicians, personalizing education, and offering emotional scaffolding for patients and caregivers.

“While empathic AI systems show promise in extending the reach of clinicians, personalizing education, and offering emotional scaffolding for patients and caregivers, their successful integration into clinical workflows remains dependent on several critical factors.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Empathic AI systems simulate empathic behaviors through pattern recognition and affective cue detection but lack lived experience, moral reasoning, and relational accountability.

“Empathic AIa systems simulate empathic behaviors through pattern recognition and affective cue detection but lack lived experience, moral reasoning, and relational accountability.”

• AssumptionRests on an assumption: True empathic understanding rooted in shared experience remains uniquely human.

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Training data may encode structural inequities or fail to account for cultural variation in emotional expression, caregiving norms, and communication styles, resulting in emotionally inappropriate or inequitable interactions.

“Training data may encode structural inequities or fail to account for cultural variation in emotional expression, caregiving norms, and communication styles, resulting in emotionally inappropriate or inequitable interactions.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Sentiment analyses or AI-generated chat summaries can be used to brief providers ahead of visits, helping them tailor communication strategies to patient needs.

“sentiment analyses or AI-generated chat summaries can be used to brief providers ahead of visits, helping them tailor communication strategies to patient needs.”

May combine two claims.

Empathic AI-mediated tools may help close gaps in emotional availability, ensure message consistency, and reduce interpersonal strain in sensitive care discussions.

“Empathic AI-mediated tools may help close gaps in emotional availability, ensure message consistency, and reduce the interpersonal strain that often accompanies sensitive care discussions.”

May combine two claims.

Empathic AI enables care that integrates clinical precision with personalized emotional support, bridging a gap that has long challenged providers operating under time and resource constraints.

“Empathic AI enables care that integrates clinical precision with personalized emotional support, bridging a gap that has long challenged providers operating under time and resource constraints.”

• AssumptionRests on an assumption: Patients with cancer face complex treatment pathways intertwined with profound emotional and existential burdens.

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

This review highlights the distinct value and complexity of applying empathic AI in oncology, where sensitivities differ markedly from other clinical contexts.

“This review highlights the distinct value and complexity of applying empathic AI in oncology, where sensitivities differ markedly from other clinical contexts.”

• AssumptionRests on an assumption: Oncology presents distinct emotional and ethical challenges, making it a uniquely high-stakes setting for the application of empathic AI.

Probably not a substantial claim: may not pass S2 (checkable).

Empathic AI holds potential to reshape the patient-to-provider relationship.

“Empathic AI also holds potential to reshape the patient-to-provider relationship.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

The pace of innovation complicates assessments of clinical readiness.

“While this pace of innovation reflects momentum in the field, it also complicates assessments of clinical readiness.”

• AssumptionRests on an assumption: AI technologies are rapidly evolving.

May combine two claims.

Models trained on nonrepresentative data may risk reinforcing racial, cultural, and linguistic biases, leading to emotionally inappropriate or harmful interactions.

“Models trained on nonrepresentative data may also risk reinforcing racial, cultural, and linguistic biases, leading to emotionally inappropriate or even harmful interactions.”

May combine two claims.

Empathic AI is positioned not as a substitute for human connection but as a decision support and interaction support that may enhance communication, accessibility, and emotional scaffolding within clinician-led care.

“This positions empathic AI to be understood not as a substitute for human connection but as a decision support and interaction support that may enhance communication, accessibility, and emotional scaffolding within clinician-led care.”

May combine two claims.

These systems help both patients and health care providers navigate sensitive conversations, ensuring that care remains patient-centered and aligned with individual values during the final stages of life.

“These systems help both patients and health care providers navigate sensitive conversations, ensuring that care remains patient-centered and aligned with individual values during the final stages of life.”