According to a 2025 study on the Expectations of healthcare AI and the role of trust, “US adults generally have low expectations of benefit from AI in healthcare, but those with higher trust in their providers and health systems are more likely to expect to benefit from AI.”

The study also found that 69.09% of respondents had low overall expectations of AI in healthcare, compared with 30.91% who had high expectations. So, while an AI system might reduce administrative workloads, many patients are less convinced that these benefits will directly improve their healthcare. Evidently, healthcare organizations evaluate AI based on operational improvements, while patients evaluate it based on their own experiences. This shows a potential “misalignment” between organizational expectations and patient expectations.

 

Increasing access to patient care

Of the three areas examined, access to healthcare generated the highest expectations, with 30.28% of respondents expecting AI to make it easier to access. For example, healthcare organizations can use AI to improve administrative processes, patient communication, scheduling, or information management, affecting how patients interact with the healthcare system.

The researchers also found that people who reported being able to get healthcare when they needed it were more likely to expect AI to provide benefits. Therefore, people who already have easy access to care may be more likely to believe AI can improve the system.

 

Improving patient-provider relationships

In the above-mentioned study, only 19.55% of respondents believed that AI would improve their relationship with their doctor. This could possibly stem from a concern that healthcare is based on human interactions, with providers offering explanations, reassurance, decision-making, and guidance. Using AI can therefore raise questions about whether technology will support those interactions or create greater distance between patients and providers.

More specifically, patients may not automatically associate greater use of technology with better communication. So, for AI to improve patient-provider relationships, patients must understand how it is being used and how healthcare providers will stay involved in their care.

 

Improving healthcare affordability

According to the research article, patients had very low expectations in terms of how AI will affect healthcare costs. Only 19.4% of respondents expected AI to make healthcare more affordable, indicating a difference between operational efficiency and patient affordability.

AI may help healthcare organizations automate certain processes, but these organizational savings do not necessarily translate directly into lower costs for patients. As a result, patients may be skeptical of claims that AI will make healthcare more affordable unless they can see evidence that the technology is producing tangible benefits.

The researchers argue that there should be less emphasis on anticipation and excitement around AI and more emphasis on “evidence and communication” about when the technology can benefit patients.

 

How trust affects patient expectations of AI

Respondents with higher trust in the healthcare system and individual providers were more likely to expect benefits from AI. Patients may be more comfortable with AI when they believe that the healthcare organization implementing it is trustworthy. The researchers also warn that “unmet or inflated” expectations may have negative consequences for trust, so healthcare organizations must be realistic when communicating about AI.

In simple terms, explaining what AI does, what it does not do, and how it is monitored and secured may give patients a more accurate understanding of its use.

 

Privacy builds trust in healthcare AI

The study identifies “transparency, privacy protections, and rigorous evaluation” as important aspects of trustworthy AI use in healthcare. Moreover, patients want to know what happens to their information when AI is used. They may also want reassurance that healthcare organizations have safeguards in place to protect sensitive information from unauthorized access.

As a result, healthcare organizations must uphold federal laws like HIPAA’s Privacy and Security Rules. More specifically, organizations must use a HIPAA compliant email solution, like Paubox, which safeguards patients’ protected health information (PHI) during transmission and at rest. Paubox email also uses advanced encryption, access controls, and audit trails to prevent unauthorized access to PHI.

 

Patient engagement is part of AI implementation

The researchers argue that “patient perspectives and expectations” need to be considered when developing healthcare AI governance. More specifically, “implementation of AI without consultation or patient engagement could further threaten the trust of patients who do not find the system or providers to be trustworthy.”

For example, a healthcare organization could conduct surveys to get feedback from patients on their preferences regarding AI in healthcare. This information can then be used to develop AI systems that are more patient-centered and consider patient needs. AI tools may improve efficiency, but patient might be more concerned about whether they can still communicate with their doctor and whether their information is protected.

Paubox email supports the use of AI and is proven to promote a trusting patient-provider relationship. Providers can use AI to segment their patient population and send personalized emails that address specific concerns, ultimately enhancing communication and building trust. Furthermore, providers can use AI to generate automated responses to common patient inquiries, allowing for quicker communication

Related: Risk of ungoverned AI use in healthcare

 

Communication can make AI more patient-centered

Healthcare organizations must explain how AI is being used and what safeguards are in place to protect patients. Communication can also give patients opportunities to ask questions about AI-assisted care.

Organizations can use HIPAA compliant emails to provide educational resources on AI in healthcare, keeping patients informed about the benefits and limitations of AI technology in their care. This transparency can help alleviate potential misconceptions patients may have about AI, building trust and confidence in the use of AI in healthcare.

For example, if a patient is worried that AI will replace their doctors, a secure email can reassure them that AI is meant to complement the work of healthcare providers, not replace them. Providing information on how AI algorithms are developed and validated can also help patients understand the reliability of AI technology in healthcare.

If a patient is worried that their personal data will be misused by AI systems, providers can use secure emails to explain the privacy protocols that protect patient information. Additionally, these emails can include information on transparency explaining how patient data is used and stored in AI systems to further build trust.

 

Connecting AI adoption with patient trust

Most respondents in the previous study did not expect AI to significantly improve affordability or their relationship with their doctor. At the same time, trust in healthcare systems and providers was strongly associated with greater expectations of AI benefits. As such, we cannot separate AI adoption from patient trust.

Healthcare organizations must prove that AI is being implemented responsibly and show that they protect the health information patients entrust to them.

HIPAA compliant emails can help patients judge healthcare AI based on whether it improves their experience, preserves their relationship with their provider, protects their information, and delivers meaningful benefits. This “evidence of patient benefit” can help build trust in healthcare AI technologies.

Healthcare organizations can also include patients in the decision-making process when implementing AI, like using secure emails to ask them for feedback on the design and functionality of AI systems.

For example, if a patient with a visual impairment is asked to provide feedback on an AI system's accessibility features, their input can lead to improvements that will also benefit other patients with similar needs. Involving patients in the decision-making process keeps AI systems user-friendly and promotes patient engagement. It can also be seen in different population groups, like older adults or non-English speaking patients, where feedback can help tailor AI systems to better meet their needs for a more personalized approach.

 

FAQs

Should patients be told when AI is used in healthcare?

Yes. Healthcare organizations should consider transparency when implementing AI systems. The research identifies transparency as a major component of trustworthy AI and suggests communicating with patients about how AI may affect their healthcare.

 

Can AI be used with protected health information (PHI)?

Yes. AI can be used in healthcare organizations that handle PHI, but these organizations must consider HIPAA’s privacy and security requirements when implementing these technologies. This includes evaluating how PHI is collected, accessed, processed, stored, and shared through an AI system.

Organizations must also check that the required safeguards are in place to protect PHI and address the risks associated with using AI. Transparency and privacy protections ultimately promote trustworthy AI use in healthcare.

 

Is AI automatically HIPAA compliant?

No. AI is not automatically HIPAA compliant. Compliance depends on how the technology is implemented, configured, used, and governed when it handles PHI.

Healthcare organizations should evaluate how an AI system collects, accesses, processes, stores, and shares PHI and implement appropriate privacy and security safeguards. They should also assess the specific AI tool and workflow to determine how HIPAA requirements apply.

For organizations that use AI to generate or personalize patient communications, the Paubox Email API can provide the secure email delivery layer. The API allows apps and healthcare workflows to programmatically send transactional emails containing PHI, with TLS 1.2+ encryption and a business associate agreement (BAA) included with every plan.

Learn more: Understanding APIs