Bangla, Hindi and Urdu speakers with visual impairments face major barriers to digital access, despite these languages holding official status. Accessibility, Language and Tech for the People (ALT) by Whose Knowledge?, is a multi-year research-in-action initiative spanning India, Bangladesh and Pakistan. ALT investigates how disability, patriarchy and class inequalities influence online participation, and who gets left out. Previous digital accessibility studies have shed light on issues such as the impact of ableist policies on technological implementation. Yet the intersection of language justice and disability in non‑English contexts across South Asia, and its deeper, long-term consequences for people with visual impairments, remains far-less studied. ALT addresses this gap by foregrounding the lived experiences of visually impaired users in their own languages and contexts, centring disability, gender and class (which is South Asia frequently translates into caste) as intersecting forces that determine whether technology expands a person’s world or quietly closes it off.As the barriers visually impaired people face are structural, the methodology had to be both collaborative and grounded. Research teams in each country were led by visually impaired researchers working in Bangla, Hindi or Urdu. They designed and conducted local‑language surveys, platform audits, in‑depth interviews and focus group discussions, treating users with visual impairments not as “beneficiaries” but as co‑researchers whose expertise shaped both the questions and the answers.Findings and thematic observationsNearly 160 participants across three countries represented a diverse array of occupations, including students, teachers, public officials, managers, writers, housewives, technologists and creators from rural and urban areas, spanning ages from people in their twenties to older adults.Access to technology is not optional, but essential. Like people without disabilities, visually impaired people depend on it for vital information, healthcare, education, employment, banking, entertainment, culture, communication and overall social well-being.Our findings reveal that users with visual impairments overwhelmingly want to use technology in their local languages, yet the systems they depend on consistently force them to fall back on English. In Pakistan, despite more than half of surveyed users preferring Urdu, poor quality Urdu support forces 89% of users to constantly switch between Urdu and English. In Bangladesh, access is severely limited with only 6% of respondents being able to access content fully in Bangla. Half of the users are forced to navigate a mix of Bangla and English. In West Bengal, 56% of visually impaired respondents reported being unable to access the internet using Bangla alone. The scale of this problem is further highlighted by an audit showing that a major Hindi news portal, Dainik Jagran, failed 45% of its automated accessibility checks.Because of these realities, English has become a need. Not because users prefer it, but because local‑language tools are broken by design and content is scarce. Available digital content often fails to capture nuanced angles, rendering it insufficient for users. In many other contexts, English has served as a tool for social equality and mobility. Here, the presence of information in English is not the problem, but the absence of the same in local languages. Those already excluded by their visual disability, are excluded twice when they lack English fluency.Exclusion is shaped by gender and class The research confirms what disability rights advocates have long argued: gender and class create layered technological exclusion, while caste, class and gender priviledges grants some users a sense of easy entitlement to access. For example, women with visual impairments in rural areas described being unable to travel for training, missing online meetings scheduled around domestic and childcare responsibilities and dropping out of education or courses when phones broke or family finances tightened.Shirin Akter, a Bangladeshi participant, noted that visually impaired women “learn only as much as is absolutely necessary” since the time, devices and support needed for deeper engagement are systematically denied to them. This dynamic gives men with visual impairments significantly more advanced technical knowledge than their female counterparts. Another woman in India, who wished to remain anonymous, never completed higher education and now relies on a basic feature phone without a screen reader. She says she has lost the motivation to obtain or learn about new tools, even as her blind brother is educated, uses a smartphone, works away from home and manages his own tasks independently. These are not individual choices. They are social inequities translated into inequitable digital access. Unstable building blocks of local-language accessibilityDespite obtaining devices and internet connectivity, the basic building blocks of accessibility remain unstable. Digital spaces are rife with content that cannot be processed by screen readers: scanned images-based PDFs, slide decks, image‑only newspapers, missing alt text, unlabelled buttons and, most importantly, unstable screen readers that stumble over local scripts or mixed‑language documents. In Bangladesh, only 6% of respondents said they could access all content in Bangla, with many noting that Bangla PDFs simply do not read correctly with available tools. Indian Bangla participants face similar challenges, even though PDF is their third-most-used document format. Manjit Kumar Ram, a PhD scholar at Jadavpur University in Kolkata, said, “90% of the online Bangla newspapers are inaccessible. On top of that, flash ads make reading harsh.”Software updates adversely affect accessibility. Ride‑hailing apps that were once usable introduce newly unlabelled buttons; entertainment apps that previously read out channel names suddenly stop after an update; TalkBack – the default screenreader for Android – and other services begin crashing on lower‑end Android phones. Users are rarely given accessible onboarding for new features and have no way to roll back to versions that worked. Emotional costs and lost futuresThese failures have profound human costs. Taiba Khatun, a Bangla literature student from Jadavpur University, came close to dropping out of her undergraduate program due to inaccessible course materials. She was also forced to choose “non-challenging” research topics because references in Bangla were not available in accessible formats. Participants also spoke of avoiding STEM (Science, Technology, Engineering, and Mathematics) subjects due to fear of inaccessible formats in which chemical or mathematical formulas are typically presented. Mental exhaustion is common among student-participants. Accessing a single image or PDF can take minutes when their peers can read the same in seconds. Some described paying human readers to audio-record textbooks for them, adding effort, time and financial costs to an already unequal experience. These realities severely limit people’s educational and employment opportunities, ultimately undermining their social and economic well-being.What do accessible futures look like?Synthesising the full body of research has helped us create various resources, tutorials, principles of practice and an accessible technology convening. These principles distill a single, non-negotiable demand from the collective voices of users: future designs must centre around and include local-language blind and low-vision users. Fulfilling this requires a fundamental shift in priorities, from profit to people. This means designing high‑quality text-to-speech and optical character recognition for Bangla, Hindi, Urdu and other local languages, building stable and accessible interfaces that survive software updates and treating affordable, offline‑capable tools as a baseline. The research also recognises that design failures are a form of discrimination, violating fundamental constitutional rights. Ironically, many government portals across all three countries are least accessible, with CAPTCHA blocks denying access to essential public services.Moving forward, states, companies and creators must invest in critical technical infrastructure necessary to bridge these gaps. This includes strengthening disability laws, enforcing compliance policies, establishing training and support hubs, translating and localising open-source tools like NVDA and developing offline OCR for regional scripts. Existing technology and systems need to be made accessible and user-friendly.We believe this research is more than just an assessment of current systemic failures. It points to an urgently necessary future, where the transformative power of technology is accessible to all, regardless of geography, income or language. For visually impaired Bangla, Hindi and Urdu users who participated in this study, gaining technology skills was profoundly empowering, restoring agency, independence and joy. Tejram Sahu, a teacher, captured this feeling perfectly by stating, “When I learned how to write, I felt that this was a rebirth or revival of my life.” Our goal should be to make that same profound sense of rebirth and independence available to everyone. This is the second part in a series on the accessibility, language and technology of visually-impaired persons. Read part one here.Subhashish Panigrahi is a nonprofit leader with a background in social entrepreneurship, digital rights, community-building, and language technology. Maari Maitreyi is a researcher and coordinator working for knowledge and epistemic justice online at Whose Knowledge? Claudia Pozo is a researcher and coordinator at Whose Knowledge? focused on knowledge and epistemic justice online. Nirmita Narasimhan works on policy research and advocacy related to intellectual property reform and technology access for persons with disabilities.