TT’S TECH TAKE 015

When Healthcare AI Doesn't Speak Your Language: Who Gets Left Behind?

If healthcare AI can answer in your language, does that mean it has actually been tested for safe use in your care?

Today’s tech. Five human questions. One healthier future.

By T. Robinson | TT’s Tech Care

When Innovation Speaks, Who Is It Speaking To? 🌍❤️

Now, THIS is the kind of healthcare technology research that immediately gets my attention!

We spend so much time celebrating the incredible possibilities of artificial intelligence in healthcare, and trust me, I'm right there cheering for the progress! From improving communication and supporting healthcare professionals to expanding access for underserved communities, the potential is exciting, especially when we consider what these advancements could mean for patients who already face significant barriers to care.

But every once in a while, research comes along that challenges us to look beyond what technology can do and ask whether we're doing enough to ensure it works responsibly for the people who need it.

And that's exactly what happened when I came across this international study!

Imagine being a patient trying to understand a new diagnosis, asking questions about medication, or preparing for an appointment. You turn to an AI-powered healthcare tool, and to your relief, it responds in your native language!

That sounds like progress, doesn't it?

Now consider something we don't always stop to ask. Has that technology actually been evaluated for clinical accuracy and safety in your language, or are we assuming that because it can communicate fluently, it can also communicate reliably?

Those are two very different things, especially in healthcare, where misunderstandings can have serious consequences.

This is where the conversation becomes bigger than artificial intelligence, because we're really talking about trust, accessibility, patient safety, and whether the future of healthcare innovation will include everyone.

The Research: 806 Studies, 62 Languages, and a Global Gap 🔬

On October 12, 2026, researchers Dean Ariel and Guy Laban published a study in Communications Health examining how clinical artificial intelligence is evaluated across languages other than English.

Their scoping review covered 806 studies involving 62 non-English languages, and the findings revealed significant differences in where multilingual healthcare AI research is being conducted.

One statistic in particular made me stop!

Of those 806 studies, just 17, approximately 2.1%, were attributed to the World Health Organization's African and South-East Asian regions, which together represent approximately 3.15 billion people.

Think about the scale of that difference!

The researchers also found that Chinese, Japanese, and Turkish accounted for more than half of the evaluations reviewed, while 14 languages spoken as a first language by more than 50 million people each had no identified clinical LLM evaluation in the databases searched.

And there's more that deserves our attention.

Only 18.6% of the reviewed studies used real-world clinical records, and none reported patient-level outcomes or patient perspectives. The research was also heavily concentrated in the Western Pacific and European WHO regions, which together accounted for more than 83% of the studies.

Now, I want to be clear about something important. These findings do not prove that multilingual healthcare AI is inherently unsafe, and the review does not capture every evaluation that may exist worldwide. It examines published research indexed in three major databases through February 2026, meaning some regional or unpublished evaluations may not have been included.

But even with those limitations, the picture raises important questions about whose languages, healthcare environments, and experiences are receiving the most research attention.

And as someone who cares deeply about communication accessibility, underserved communities, and the human side of healthcare, I believe those questions deserve much more than a passing conversation!

💻❤️ TT’S TOP 5

1️⃣ What did the researchers actually discover?

What stood out to me wasn't simply the number of studies reviewed, but how unevenly those evaluations were distributed across languages and regions.

The research suggests that clinical AI evaluations remain concentrated in languages with stronger research representation, while many others receive considerably less attention. That imbalance matters because demonstrating strong performance in one language doesn't automatically establish that a system will perform equally well in another.

And here's where I believe we need to challenge how we define innovation!

It's easy to celebrate an AI system that can respond in dozens of languages, but I want to know how those responses were evaluated, whether the medical information was accurate, and whether the system was tested in environments that reflect the communities expected to use it.

A patient asking about symptoms in a rural clinic in Kenya may have different linguistic, cultural, and healthcare circumstances than someone using the same technology in a major hospital in Japan or the United States.

Neither patient deserves less attention to safety or quality!

For me, the takeaway is that healthcare AI should be evaluated with the same seriousness we expect from other technologies influencing patient care, while recognizing that language and local clinical context can change how a system performs.

Innovation shouldn't just be measured by how many languages technology can speak, but by how responsibly we evaluate its ability to support the people speaking them!

2️⃣ Why does language matter beyond simply translating words?

Oh, this question is especially close to my heart! ❤️

After years of working in healthcare across hospitals, clinics, and different patient-facing environments, I've learned that effective communication is about so much more than exchanging information.

It's about understanding what someone is experiencing, recognizing concerns they may struggle to explain, and making sure they feel comfortable enough to ask questions without embarrassment or fear of being misunderstood.

My own experiences navigating communication accessibility have made that lesson even more meaningful.

As someone who stutters, I understand the importance of being given the time, patience, and opportunity to communicate without someone making assumptions about what I'm trying to say. That experience isn't identical to facing language barriers in healthcare, but it has shaped how deeply I value communication access and why I advocate for it so passionately.

Now imagine bringing artificial intelligence into that experience.

Medical terminology doesn't always translate neatly between languages, cultural differences can influence how patients describe symptoms, and certain expressions may carry meanings that are difficult for technology to interpret without appropriate context.

An AI system might produce a response that sounds perfectly fluent while missing a medically important detail.

That's why I don't believe translation alone should be our standard for success!

I want healthcare technology to support meaningful understanding, not simply generate words that sound convincing. Patients deserve information that is clinically appropriate, understandable, and responsive to their circumstances.

Because genuine accessibility isn't just about making information available. It's about giving people a meaningful opportunity to understand, participate, and be heard!

3️⃣ Who risks being left behind if these gaps continue?

This is where my passion for rural healthcare and underserved populations really comes into focus!

Whenever I read research about gaps in healthcare technology, I find myself thinking about the patients and healthcare professionals who are already doing so much with limited resources.

I'm thinking about communities where specialty care may be hours away, facilities where staffing shortages affect access to services, and patients who already face challenges related to language, transportation, geography, or digital access.

Now imagine introducing AI into those environments without sufficient evidence about how well it performs in the languages spoken by the people relying on it.

Technology that was intended to close healthcare gaps could unintentionally introduce new ones, and that's something I never want us to overlook.

The limited evaluation coverage identified across the African and South-East Asian WHO regions is particularly concerning because these regions represent enormous populations with diverse languages and healthcare needs.

And while this study focuses on international research, the implications absolutely reach us here in the United States!

We have multilingual communities, rural populations, and patients who face communication barriers every single day. Whether someone is seeking care in a small American town or a community thousands of miles away, the importance of understanding their healthcare information doesn't change.

What concerns me most is the possibility that communities already experiencing healthcare inequities could also become underrepresented in the research shaping future technology.

I want us to build a future where the communities facing the greatest barriers to care are included in innovation from the beginning, not considered after the technology has already been developed!

4️⃣ Does this mean multilingual healthcare AI is unsafe?

Not necessarily, and I think this distinction is incredibly important!

The study identified substantial gaps in published evaluations of multilingual clinical AI, but it did not establish that every AI tool operating in an underrepresented language is unsafe.

A lack of published evidence isn't the same as demonstrated harm. However, it also shouldn't be mistaken for evidence that a system is ready for every clinical environment.

And when we're discussing healthcare, that uncertainty deserves serious consideration!

One finding that caught my attention was that only about one-quarter of the reviewed studies directly compared performance between English and non-English languages.

The researchers also identified heavy reliance on medical examination questions and simulated clinical scenarios rather than real-world patient care.

Now, don't get me wrong, performing well on medical questions can be useful when evaluating AI! But answering a structured question isn't necessarily the same as navigating a conversation with a worried patient who may describe symptoms differently, leave out important details, or need additional explanation.

Healthcare is rarely as straightforward as choosing the correct answer on an examination.

That's why I want organizations to look beyond impressive demonstrations and ask how these systems perform in real clinical workflows, how errors are identified, and what safeguards exist when technology produces an inaccurate or inappropriate response.

My enthusiasm for artificial intelligence hasn't changed one bit! I still see tremendous potential, especially when these systems are designed to support healthcare professionals rather than replace human judgment.

But responsible innovation requires us to understand the limitations alongside the possibilities.

We should never confuse an AI system's ability to sound confident with evidence that its medical information is reliable, particularly when a patient's health and well-being are involved!

5️⃣ What needs to change to make healthcare AI more equitable worldwide?

Now THIS is where I want to turn concern into action! 🙌🏽

One encouraging part of the research is that the authors didn't simply identify disparities. They also proposed a reporting framework called LINGUA, which stands for Language Inclusivity Norms for Generative AI use in Clinical Assessment.

The proposed framework emphasizes clearer reporting around translation methods, cultural adaptation, comparisons with English-language performance, and results presented separately for each language.

It's still a proposal that requires additional validation before formal adoption, but I appreciate the direction because transparency is essential when we're talking about healthcare technology.

Personally, I want to see developers, researchers, healthcare leaders, and policymakers take an even broader approach!

Start by involving healthcare professionals who actually work in the communities where these technologies may be used. Their knowledge of local languages, clinical practices, and patient needs can help identify issues that might never appear in a controlled testing environment.

And please, let's include patients in these conversations!

I cannot emphasize that enough.

If we're developing technology intended to improve the patient experience, then patients should have opportunities to influence how that technology is designed, evaluated, and introduced into care.

I also want to see more testing involving real clinical settings, stronger language-specific performance reporting, and meaningful investment in research that includes communities historically underrepresented in healthcare technology development.

Not every region has access to the same resources, but that shouldn't mean certain populations remain invisible when research priorities are established.

Above all, I believe we need to stop treating health equity as something we address after innovation happens.

Equity should help shape the innovation itself!

Because when technology is developed with people rather than simply for them, we create opportunities for solutions that are more thoughtful, more inclusive, and better aligned with the realities of healthcare.

Beyond the Numbers: The Human Side of Healthcare AI ❤️

There's one additional finding from this research that I really want to spend time discussing, because it speaks directly to why TT's Tech Care exists.

None of the 806 studies reviewed reported patient-level outcomes or patient perspectives.

Now, that doesn't mean patient-focused multilingual AI research doesn't exist elsewhere, and I certainly don't want to suggest that researchers aren't doing important work to understand patient experiences.

But within the studies examined, this was a meaningful gap!

We can measure how accurately an AI system answers medical questions, compare its performance across languages, and evaluate technical capabilities, but I also want to understand what happens when real people encounter these tools during real healthcare experiences.

Does the technology help someone better understand their diagnosis? Does it make an appointment less intimidating or give a patient greater confidence in asking questions? Can it support healthcare professionals in communicating more clearly without introducing unnecessary confusion?

And perhaps most importantly, does it contribute to better, safer, and more equitable patient care?

Those are the outcomes I want us to keep working toward!

Throughout my healthcare career, I've encountered patients navigating moments filled with uncertainty, frustration, fear, and hope. Sometimes the most meaningful part of an interaction wasn't a complicated procedure or an impressive piece of equipment, but taking the time to listen and making sure someone understood what came next.

That human connection matters just as much today as it did before artificial intelligence became part of our everyday healthcare conversations.

And while AI may offer remarkable tools to support communication, it shouldn't replace the responsibility we have to make patients feel respected, included, and understood.

Patients are not simply data points in an evaluation. They're people trusting healthcare systems with some of the most important parts of their lives, and they deserve technology developed with that responsibility in mind!

💬 TT’S TAKE

For me, language access is healthcare access! ❤️

And the more I explore the future of artificial intelligence in healthcare, the stronger that belief becomes.

I genuinely LOVE seeing what technology can make possible! I love learning about new research, discovering creative solutions, and imagining how innovation could help healthcare professionals spend more time doing what matters most: listening, connecting, educating, and caring for their patients.

Especially when those advancements could help bridge gaps for rural communities, underserved populations, and people who have historically struggled to access the healthcare resources they need!

But even with all that excitement, I never want us to lose sight of something essential.

Behind every technological advancement is a person who may eventually depend on it.

I think about a mother trying to understand her child's symptoms, an older adult learning how to manage a new medication, or a patient in a rural community searching for trustworthy information before making a long trip to a medical appointment.

I think about the healthcare professional doing everything possible to communicate clearly with someone whose language or cultural experiences differ from their own.

And I think about the patients who already face barriers to being understood, whether those barriers involve language, disability, geography, or access to technology.

These are the people I want us to remember when we celebrate the future of healthcare AI!

Because while I absolutely believe artificial intelligence has the potential to transform healthcare, I don't want progress to be defined solely by technological capability.

I want it to be reflected in the way people experience care.

Can patients ask questions and receive information they can understand? Can healthcare professionals trust that the tools supporting them have been appropriately evaluated? Are underserved communities being included in research, rather than being expected to accept solutions developed without their perspectives?

And are we making healthcare more accessible without sacrificing the compassion and human connection that make it meaningful?

Those are the questions that continue to inspire my work.

Because what good is advancing healthcare if we forget the people we're advancing it for?

That belief is at the heart of TT's Tech Care, and it's why I will always advocate for innovation that keeps humanity at the center.

Not just for patients who speak the most widely studied languages or live near major medical centers, but for the people whose voices are too often missing from conversations about the future of healthcare.

I want us to celebrate every meaningful technological breakthrough while continuing to challenge ourselves to make that progress more inclusive, responsible, and accessible!

Innovation becomes meaningful when technology creates more room for humanity, no matter what language that humanity speaks! 🌍❤️

And THAT is a future of healthcare worth working toward!

🎤 Now I Want to Hear From You!

If healthcare AI is being introduced into patient-facing environments, should organizations be required to demonstrate language-specific safety and performance before using it in patient care?

And what can researchers, healthcare leaders, and technology developers do to ensure that underserved communities have a meaningful voice in shaping the next generation of healthcare AI?

I'd love to hear your perspective, because advancing healthcare responsibly is a conversation worth having across every language, community, and border! ❤️

🔗 Research

Ariel, D., & Laban, G. (2026). Linguistic inequity in clinical AI evaluation: A scoping review of 806 studies across 62 languages. Communications Health, 1, Article 39.

https://www.nature.com/articles/s44528-026-00042-1

Previous
Previous

TT’S TECH TAKE 016

Next
Next

TT’S TECH TAKE 014