What a New JGIM Study Tells Us About AI and Human Medical Interpreters
TL;DR
A new study in the Journal of General Internal Medicine identifies 11 ways medical interpreters support patient safety beyond spoken language. Julie Mills, Boostlingo’s Chief Nursing Informatics Officer, shares her perspective on where AI interpreting can expand language access, why human interpreters remain essential, and how quality measurement and governance can support responsible AI adoption in healthcare.
Medical interpreters do much more than interpret conversations between patients and healthcare providers. They help recognize confusion, navigate cultural differences, and catch misunderstandings that could affect patient safety. But can AI do the same?
A new study published October 5, 2026, in the Journal of General Internal Medicine (JGIM) titled The Role of Interpreters in Supporting Resilience and Implications for Artificial Intelligence explores 11 ways medical interpreters support patient safety beyond interpreting spoken language. The findings highlight responsibilities that today’s AI interpreting tools don’t yet fully replicate, while raising important questions about how AI could fit into healthcare communication.
To better understand what these findings mean for healthcare organizations, we spoke with Julie Mills, Chief Nursing Informatics Officer (CNIO) at Boostlingo, about where AI could expand language access and why human interpreters remain essential to patient care.
What Do Medical Interpreters Do Beyond Interpreting?
The researchers examined the experiences of 17 current and former medical interpreters, including 12 who participated in three focus groups and five who helped review and contextualize the findings.
The study focused on ambulatory diagnostic care, where patients seek evaluation for new or worsening symptoms.
So, what does a medical interpreter actually do beyond interpreting words? According to the study, quite a bit:
| Category | Key findings |
|---|---|
| Respond | Interpreters manage communication flow, strengthen the patient’s voice, provide broader support, and participate in pre-appointment briefings. |
| Monitor | Interpreters recognize patient confusion, emotional cues, team dynamics, and technology-related communication problems. |
| Learn | Interpreters use cultural knowledge and previous experiences to recognize misunderstandings and improve communication. |
| Anticipate | Interpreters identify potential safety risks and opportunities to improve care processes before problems develop. |
Imagine a patient finishing an appointment and telling their doctor they have no questions. But once the doctor leaves, the patient turns to the interpreter and admits they’re confused about their treatment.
That’s one of the situations interpreters described in the study. It shows how interpreters can recognize communication problems that might otherwise go unnoticed. Others described using cultural knowledge to identify misunderstandings that could affect a patient’s willingness to follow medical recommendations.
The researchers emphasized that AI interpreting systems must account for these additional responsibilities as healthcare organizations introduce automated language services.
The study was qualitative and based on interpreters’ reported experiences. It did not directly compare AI and human interpreting performance or measure differences in clinical outcomes.
Where AI Interpreting Could Expand Healthcare Communication
Although the study identifies responsibilities that current AI systems do not fully replicate, it does not rule out AI interpreting in healthcare.
Mills sees an opportunity in the quick, everyday conversations that often happen without an interpreter.
“There are countless quick conversations that happen on the perimeter of clinical care, where on-demand human interpreters are the only option without AI. These are often simple exchanges of a few back and forths without any clinical context, hence the reason we see people pull Google Translate out of their pockets, both patients and providers. AI is already being used, but there’s an opportunity to use it within the realm of compliance, with proper quality measurement and safety guardrails.”
“Check in, routine calls and standard follow up are all quick moments where little to none of the conversation is clinical terminology or diagnoses, making AI an efficient and meaningful collaborator. More importantly, having a human one phrase or tap away ensures you can transition to a more supportive interpretation session at any time.”
These everyday interactions offer an opportunity to expand language access, but healthcare organizations still need to consider when professional interpreting is necessary.
Why Human Interpreters Still Matter
Healthcare conversations can become complex quickly. A routine appointment might lead to questions about a diagnosis, treatment options, or concerns a patient has trouble expressing.
In these situations, professional interpreters provide more than language assistance. They can recognize when someone needs clarification, respond to cultural differences, and help patients communicate their concerns.
That’s why Julie Mills emphasizes keeping human interpreters readily accessible as healthcare organizations introduce AI. The ability to transition from AI to a professional interpreter gives healthcare teams additional support when conversations become more complex or sensitive.
Choosing the appropriate interpreting method ultimately depends on the situation and the potential consequences of communication errors.
Why Quality Measurement and Governance Matter
Getting the words right is important, but how do we know whether an AI interpreter is supporting effective communication?
Recognizing confusion, responding to clarification requests, and managing conversational flow are all part of effective communication, and traditional translation accuracy measurements can miss them.
For Julie Mills, that means finding ways to measure these additional aspects of interpreting quality.
“The first step is the ability to objectively measure cues, clarifications or conversational challenges, so that they can be understood and evaluated.”
Boostlingo Assure is one example. Mills says it addresses a limit of human review:
“Human QA only supports limited insight due to the time and training required to QA effectively, but the market now holds an objective quality-assurance tool, built and trained by industry leading interpreting experts, called Assure.”
Setting Clear Rules for AI Interpreting
But measuring quality is only one part of using AI responsibly. Healthcare organizations also need clear policies governing when, where, and by whom AI interpreting can be used.
Julie believes these safeguards should be built directly into the technology, rather than relying entirely on staff to make those decisions.
“Technology owning the governance and control, only allowing certain departments or users to use it in accordance with policy, is a great first step that Boostlingo has introduced.”
Healthcare organizations should also review how AI is being used, monitor adherence to internal policies, and adjust access as their needs evolve.
AI Medical Interpreter and Assure
Boostlingo’s AI Medical Interpreter combines real-time AI interpreting with access to professional human interpreters. It has built-in administrative controls that help organizations manage where and how AI is used, while Boostlingo Assure supports quality measurement and improvement across more than interpreting, including professionalism, conversational flow, and technical performance.
As healthcare organizations explore AI interpreting, the study highlights the importance of expanding language access while keeping patient safety and human expertise at the center of care. At Boostlingo, we’re working toward that goal with Boostlingo AI interpreter and want to highlight useful guidance for healthcare organizations on how AI and human interpreters can work together.
You can read the full study in the Journal of General Internal Medicine
or explore Boostlingo AI Interpreter to learn more about our approach.