Be My AI Gives a Million Blind Users Instant Visual Descriptions

Imagine wanting to know whether the milk in your fridge has expired, what color shirt you are holding, or whether the crosswalk signal has changed. For the estimated 340 million people worldwide who are blind or have low vision, dozens of small visual questions like these arise every single day. An app called Be My Eyes, together with its AI assistant Be My AI, now answers many of them in seconds, for free, and as of March 2026 it serves a community of more than 1 million blind and low vision users supported by over 10 million sighted volunteers.

The significance is straightforward. Everyday visual information that sighted people process without a thought has always required asking another person, waiting, or going without. By combining a massive volunteer network with a vision language model that never sleeps, Be My Eyes has turned visual assistance into something available on demand, in more than 150 countries and in over 180 languages.

From Volunteer Video Calls to AI in Seconds

Hans Jørgen Wiberg, who is himself losing his vision, founded Be My Eyes in Denmark in 2015. The original idea was beautifully simple. A blind user opens the app, points the phone camera at whatever they need help with, and a sighted volunteer somewhere in the world joins a live video call to describe the scene, read a label, or walk them through a task. The service grew organically into the largest support platform of its kind, and it remains free for users.

Human volunteers, however, take time to connect, and some questions feel too small to bother another person with. In 2023 Be My Eyes began working with OpenAI to build an AI alternative on top of GPT-4, the first version of the model capable of analyzing images and generating detailed natural language descriptions. The result, Be My AI, lets a user take a photo and receive a rich description in seconds, then ask follow-up questions by typing or voice. Is the yogurt expired? What does this painting depict? Are these pills the right ones? The model answers conversationally, and it never needs to be asked politely.

The company did not ship this quickly. Be My AI spent seven months in beta with feedback from more than 19,000 blind testers before its public launch in September 2023. That community-driven refinement mattered, because blind users are experts in what makes a description useful, and they pushed the product toward the detail and reliability the community actually needed.

Proof in the Enterprise: The Microsoft Pilot

The clearest evidence of what AI descriptions can do came from an unexpected place, customer support. In 2023 Microsoft integrated Be My AI into its Disability Answer Desk so blind customers could resolve technical issues through image and text conversation instead of phone queues. The results were striking. Be My AI resolved inquiries in an average of four minutes, less than half the average call time with human agents, and only 10 percent of users chose to escalate to a person afterward.

That pattern, AI handling routine visual questions instantly while humans remain one tap away, has become the platform's design philosophy. Be My AI is not positioned as a replacement for white canes, guide dogs, or human volunteers. It is a first line of assistance for the long tail of small questions, with volunteers still available for judgment calls, complex scenes, and anything the AI handles poorly.

Scale, Reach, and the Numbers Behind It

The usage numbers tell the story of a service that has become infrastructure for its community. In 2024 Be My Eyes handled more than 43 million assistance requests in a single year, and the company reported threefold year-over-year growth in monthly AI sessions. In January 2025 the community stood at roughly 750,000 blind and low vision users and 8.3 million volunteers. By March 2026, announced at the CSUN Assistive Technology Conference, those figures had grown past 1 million users and 10 million volunteers.

Accessibility shaped the engineering choices. Be My AI's written responses are available in dozens of languages, which also made the tool newly useful for deafblind users who pair it with braille displays, since text descriptions can be read where audio or a live video call cannot.

The Honest Limits of AI Description

Be My Eyes is candid that AI vision has real limits. Vision language models can hallucinate, describing objects that are not present or misreading text, and errors matter more when a user cannot double-check the answer visually. For safety-critical moments like navigation, the company explicitly directs users to trusted tools and human help. The World Health Organization estimates at least 2.2 billion people have a near or distance vision impairment, and roughly half of those cases are preventable or unaddressed, a reminder that technology complements rather than replaces eye care and accessibility in the built environment.

The hybrid model is the point. AI answers the easy questions instantly and privately. Volunteers answer the hard ones. In March 2026 the company doubled down on the long game by launching the Be My Eyes Foundation, an initiative to accelerate free accessibility technology for the hundreds of millions of blind and low vision people still unserved.

What a Photo Is Worth

The most common feedback the company receives is not about model benchmarks. It is about independence: checking an expiration date without waiting for a family member to come home, choosing an outfit with confidence, or simply knowing what is in a photo a friend just sent. A million users and ten million volunteers later, the founding insight holds. Most people want to help, and when AI removes the friction of asking, the world becomes readable one photo at a time.