Opinion

AI in NHS Care Is Failing Patients Who Need Language Support

The NHS is racing to become the world's most AI-enabled health system. But as Healthwatch's latest report reveals, many patients, particularly those with communication needs, are being left behind. The missing piece isn't more AI. It's the right language services provider working alongside it.

A recent Healthwatch analysis on AI in NHS care painted a picture that will feel familiar to anyone working in healthcare communication: patients struggling to get understood, AI chatbots that can't handle real human complexity, and vulnerable people falling through digital gaps.

The findings are striking. Patients are turning to generic AI tools like ChatGPT out of desperation: because they can't get timely GP appointments, NHS 111 is too slow, and/or waitlists stretch into weeks. And while some find the personalised advice helpful, others receive potentially dangerous misinformation: one patient was told they likely had shingles when they actually had Lyme disease, a condition requiring urgent treatment.

But perhaps the most concerning part of the report is what it reveals about accessibility. People registered as blind found AI chatbots incompatible with screen-reading software. Patients with autism or learning difficulties reported that AI systems simply couldn't understand them. And for the millions of people across the UK who don't speak English as a first language? The gap is even wider.

This is where a specialist language services provider becomes essential: not as a replacement for NHS AI ambitions, but as the critical infrastructure that makes those ambitions actually work for everyone.

The communication gap NHS AI can't close on its own

The Healthwatch report highlights that AI tools used for NHS administration, including booking appointments, managing prescriptions, and navigating care pathways, consistently fail patients with communication needs. WhatsApp AI receptionists are difficult to navigate. Medication management systems can't read detailed patient notes. And when these tools don't understand a patient, there's no fallback.

For patients who speak a language other than English, the challenge compounds. A generic AI chatbot trained primarily on English-language data is not equipped to deliver accurate, safe health advice in Urdu, Polish, Arabic, Somali, or any of the hundreds of languages spoken across the UK. The risks of mistranslation in a healthcare context aren't just inconvenient. They can be life-threatening.

This is precisely the gap that a professional language services provider fills. At Tolk, we've spent nearly a decade building the infrastructure that connects patients and healthcare professionals across language barriers, with both human expertise and purpose-built AI that's designed for the public sector.

Why healthcare needs a purpose-built language solution

Generic AI tools like ChatGPT aren't built for healthcare communication. They lack the domain-specific training, the safety protocols, and the accountability frameworks that clinical environments demand. When the Healthwatch report describes a nurse testing an AI tool and finding it recommending unnecessary blood tests, that's not just a quirk. It's a systemic design problem.

Healthcare language support needs to be different. It needs to be clinically aware, data-secure, and available instantly. That's why we developed PretPal: a secure, real-time AI speech-to-speech translation tool built specifically for the public sector.

PretPal isn't a general-purpose chatbot. It's a pocket-sized AI interpreter that delivers fluent, real-time voice-to-voice translation across more than 35 quality-assured languages, every one tested and approved by professional linguists. It's designed for the exact scenarios the NHS deals with daily: a patient arriving at A&E who speaks only Tigrinya, a pharmacist explaining dosage instructions to a Polish-speaking parent, a mental health assessment where nuance and accuracy are paramount.

And critically, when AI isn't enough, when the conversation is too complex, too sensitive, or too high-stakes, PretPal connects to a certified human interpreter in under eight seconds, directly within the same app. That seamless handoff between AI and human expertise is something no generic chatbot can offer.

Data security that meets public sector standards

One of the underreported concerns in the NHS AI conversation is data security. When patients type their symptoms, medical history, and personal details into ChatGPT, that data is processed by a commercial platform with no obligation to meet NHS information governance standards.

As a language services provider built for the public sector, Tolk takes a fundamentally different approach. PretPal encrypts all data, processes it within the EU in full GDPR compliance, and deletes all conversation data after each session ends. We're ISO 27001 certified for information security. No patient data is stored permanently, and healthcare organisations retain full control.

For NHS trusts navigating the complexities of the 10 Year Health Plan's AI ambitions, this isn't a nice-to-have. It's a baseline requirement.

Accessibility by design, not as an afterthought

The Healthwatch report makes clear that AI accessibility in the NHS is failing vulnerable groups: people with sensory impairments, cognitive disabilities, learning difficulties, and mental health conditions. One patient summarised it plainly: "All I want to do is order a prescription, but because of my learning difficulties and speech issues, this AI system won't understand me like a real human."

As a language services provider covering over 210 languages and dialects, Tolk is built around the principle that communication access is a right, not a feature. Our platform supports on-site, telephone, and video interpreting alongside AI translation, meaning there's always a pathway to understanding, regardless of a patient's digital literacy, cognitive ability, or language background.

This flexibility matters. Not every patient interaction needs a full human interpreter, and not every interaction can be safely handled by AI alone. The most effective NHS language strategy uses both, matched intelligently to the situation. That is exactly what Tolk's booking system does, automatically matching needs with the right language specialist from a network of thousands of professionals.

What NHS Trusts should do next

The Healthwatch findings are a wake-up call, but they're not a reason to abandon AI in the NHS. They're a reason to implement it properly, with the right partners, the right safeguards, and the right commitment to accessibility.

For any NHS trust or healthcare provider looking to integrate AI responsibly, partnering with a specialist language services provider should be a foundational step. Language access isn't a peripheral concern in healthcare. It's central to patient safety, clinical accuracy, and equitable care.

If you're exploring how AI-powered language solutions can support your patients and staff, we'd welcome the conversation. Book a demo of PretPal or get in touch with our team to discuss how Tolk can support your NHS language services strategy.