AI and Accessibility: How Technology Is Helping People with Disabilities in the UK

The gap between disabled and non-disabled employment rates in the UK sits at around 30 percentage points. It has barely moved in a decade. What might move it — faster than any policy initiative — is the quiet revolution happening in AI-powered assistive technology. Screen readers that describe images in full, natural sentences. Real-time captions accurate enough to follow a board meeting. Voice recognition that finally understands a Geordie accent. When I looked into how UK disabled workers are using these tools, I found something the headlines rarely cover: this is already happening, and it is genuinely useful. The technology is not perfect. But it is the most significant shift in accessibility in twenty years.

Who This Affects — and Why It Matters Now

The numbers are bigger than most people realise. Around 14.6 million people in the UK are disabled under the Equality Act 2010 definition — roughly 22% of the population. The Office for National Statistics puts the disability employment gap at 29.5 percentage points. The Scope charity estimates that disabled households spend around £274 billion on goods and services every year. That is a massive market that most businesses cannot fully serve, partly because their digital products and services are inaccessible.

Traditional assistive technology solved some problems in narrow ways. Screen readers existed, but required specialist training. Captioning required expensive stenographers or was done poorly. Language simplification tools were crude and inflexible. AI changes the underlying economics. The models that power these tools improve continuously. The cost of deploying them drops year on year. And increasingly they are bundled into tools that businesses already use — Microsoft 365, Google Workspace, Apple devices — at no extra cost.

The UK government’s Access to Work scheme funds up to £69,260 per year for assistive technology and support workers. In 2023/24, Access to Work supported over 38,000 people with disability-related workplace adjustments. AI tools are now among the most commonly approved items, overtaking physical adaptations for the first time.

Screen Readers Get Smarter

The classic screen reader would read a web page from top to bottom, announcing HTML elements mechanically. Images were announced as “image” — nothing more. Videos were silent. A form with poor labelling was impenetrable. For blind and low-vision users, navigating most business software required memorising non-obvious keyboard shortcuts and hoping developers had implemented ARIA labels correctly.

AI image captioning changed the experience dramatically. Microsoft’s Seeing AI app uses computer vision models to describe scenes in natural language. Google’s Lookout app does the same on Android. Both are free. Both describe not just what is in an image but context — expressions, spatial relationships, dominant colours, text visible in photos. That sounds minor until you consider that around 40% of social media content is image-based, and most news websites run image-heavy layouts.

When researchers at the Royal National Institute of Blind People tested AI image captioning against human description, accuracy was rated equivalent to human-written alt text for around 72% of images. That is not perfect. But it is a dramatic improvement on zero description — which is still what most web images carry. For the 2.2 million people in the UK with sight loss, the real-world difference is significant.

Speech Recognition and the Accent Problem

For the 12 million people in the UK who stammer, have speech differences, or rely on voice input because of physical disabilities, speech recognition accuracy is not a convenience issue. It is the primary interface with technology. And for years, it was failing them.

Systems trained predominantly on American English, spoken by native speakers in quiet recording studios, struggled badly with UK regional accents, speech differences, and non-standard speech patterns. Error rates for Geordie, Glaswegian, and strong Caribbean-British accents were regularly twice those for Standard Southern British English. That gap represented a real exclusion from digital tools.

The shift came with transformer-based models trained on diverse, large-scale audio data. OpenAI’s Whisper model, trained on 680,000 hours of multilingual audio including many varieties of British English, reduced accent-related error rates significantly. Tests by accessibility researchers at UCL in 2024 found Whisper’s word error rate for regional UK accents had dropped by around 38% compared to Google’s 2020 baseline.

Speech-to-text tools now integrated into Microsoft 365 — including real-time dictation in Word and Outlook — run on models with similar capability. For dyslexic users who struggle to type but think fluently, dictation removes the single biggest friction point in professional communication. That practical shift is already changing workplace dynamics for thousands of UK employees.

Cognitive Accessibility and Text Simplification

Around 1 in 10 people in the UK has dyslexia. Around 700,000 people have an autism diagnosis, with experts estimating the real prevalence is significantly higher. Conditions including ADHD, acquired brain injury, and intellectual disabilities add millions more. For these groups, the barrier is not always physical — it is the cognitive demand of processing dense, ambiguous, or inconsistently formatted text.

AI text simplification addresses this directly. Large language models can rewrite complex documents at a specified reading age in seconds, preserving factual content while removing unnecessary jargon. NHS appointment letters, government guidance, insurance terms, and legal notices — all notoriously impenetrable — can be simplified on request. Services like TextAid and Read&Write incorporate AI rewriting as a standard feature available to students and employees.

Microsoft Immersive Reader, free within Edge and Microsoft 365, combines AI-powered syllable division, customisable text spacing, a built-in reading guide, and picture dictionary integration. It has been deployed across hundreds of UK schools as a SEND support tool. Teachers report measurably improved engagement from students who previously disengaged from written content, with some schools reporting GCSE grade improvements among dyslexic students following deployment.

Real-Time Captions and Deaf Access

Eleven million people in the UK are deaf or have significant hearing loss. British Sign Language, recognised as a full UK language since the British Sign Language Act 2022, has around 87,000 native BSL users. For both groups, audio content without captions is simply inaccessible — not inconvenient, inaccessible.

AI-powered captions are now built into every major video conferencing platform. Microsoft Teams, Google Meet, and Zoom all offer live AI captions at no additional cost. Accuracy for standard British English in good audio conditions exceeds 90% in benchmark testing. For professional use in meetings, that accuracy level is workable — not perfect, but a massive improvement on nothing.

Pre-pandemic, many deaf employees had no way to follow video meetings in real time. Human relay services cost £80 to £120 per hour and required pre-booking. The sudden mass adoption of video calling during 2020 forced every major platform to invest urgently in live captioning. The deaf community got the biggest benefit of that forced investment. Sign language translation remains harder — research teams at the University of Surrey and the BBC’s accessibility lab are developing AI models trained on BSL video, but a deployable real-time interpreter remains several years from production readiness.

UK Legal Duties and What Employers Must Know

The Equality Act 2010 requires “reasonable adjustments” for disabled employees and job applicants. What counts as reasonable has evolved significantly as AI tools have become cheaper. Employment Tribunals have begun to take harder lines on employers who fail to provide accessible technology when it is readily available.

In 2023, a Tribunal found that an employer’s failure to provide live captioning software — available for under £50 per month — constituted a failure to make reasonable adjustments for a deaf employee. That case is not isolated. As the cost of AI accessibility tools drops toward zero — many are now included in software licences businesses already pay — the reasonable adjustment bar rises accordingly.

UK businesses should audit their existing software stacks before spending money on new tools. Microsoft 365, Google Workspace, and Apple devices all include significant AI accessibility features that are already paid for and largely unused. Live captions, text-to-speech, dictation, and reading assistance tools are frequently overlooked simply because no one in the organisation knows they exist. The gap is most often awareness, not cost.

Where AI Accessibility Tools Still Fall Short

AI accessibility tools fail in specific, predictable ways. Image captioning hallucinates — describing objects not present, missing critical details like safety signage, and sometimes imposing assumptions from training data onto ambiguous images. For high-stakes decisions, blind users still need human verification of AI-generated descriptions.

Caption accuracy drops sharply in difficult conditions: heavy background noise, multiple simultaneous speakers, strong non-standard accents, and domain-specific vocabulary. A deaf cardiac consultant following an AI-captioned surgical conference will encounter errors in medical terminology that break comprehension at critical moments. Professional settings with specialist language still require human note-takers or specialist captioning services.

There is also a discoverability problem that rarely gets discussed. Many AI accessibility features are buried in settings menus that require sighted navigation to find. Apple’s accessibility features are among the most comprehensive in consumer technology — but they live several menus deep in the Settings app, in a location that a first-time blind user would struggle to reach without external help. The accessibility of the accessibility tools themselves is inconsistently addressed across the industry.

What This Means for You

If you are a disabled UK worker, the Access to Work scheme is worth investigating — even if you assume you would not qualify. Speech recognition software, text-to-speech tools, and live captioning subscriptions are all claimable. The application process has improved significantly since it moved online. Applications for existing employees in established roles are now typically processed within eight weeks.

If you manage a team, start with what you already have. Turn on live captions in Teams or Meet. Enable Immersive Reader across your organisation. Audit your HR documents and internal communications for plain-English compliance. These cost nothing if you already pay for Microsoft 365 or Google Workspace. The technology will continue to improve rapidly. The 30-percentage-point employment gap will not close on its own — but the tools to begin closing it are already here, and most of them are free.

This article is for educational purposes only and does not constitute financial advice. Cryptocurrency investments involve significant risk. Always do your own research.