Accessibility & Technology · Article
What the Internet Sounds Like When You Can't See It
A blind screen reader user describes what the internet actually sounds like in 2026: what works, what still breaks, CAPTCHAs, AI image descriptions, and what builders can do better.
Published · Updated · 16 min read
The first thing I hear most mornings is not an alarm. It is a voice.
It is fast, faster than most people can follow, and it reads me the lock screen before I have fully sat up: the time, the battery, a stack of notifications that pile up overnight when you live in the Philippines and half the people and services you deal with run on American hours. I swipe with one finger and the voice moves to the next item. I double-tap and it opens. If an app was built well, the voice tells me exactly what is in front of me. If it was not, I hear something like "button," and then "button" again, and then "image," and I start guessing.
That voice is a screen reader. On my phone it is VoiceOver, the one Apple builds into every iPhone. On a computer it might be NVDA, which is free, or JAWS, which is not. I have been blind since birth, from a condition called Leber congenital amaurosis, so I have never used the internet any other way. I do not experience the web as a page. I experience it as a sequence. A line, then the next line. A heading, a link, a form field, a button. Everything a sighted person takes in at a glance, I take in one piece at a time, in the order the developer happened to put it in the code.
People sometimes ask me what the internet is like for a blind person in 2026, usually with the assumption that technology has solved most of it by now. Some of it has. A lot of it has not. And the parts that have changed are strange in ways I do not think most sighted people have noticed yet. So this is my attempt to describe it honestly: what works, what still breaks, and what it feels like to live in the middle of that.
A web built in a particular order
The easiest way I can explain a screen reader is this. Imagine someone reading a website to you over the phone, but they are only allowed to read what is written in the code, and they cannot skip ahead unless the page gives them a reason to.
When a page is built well, it gives me reasons. It has real headings, so I can press a single key and jump from section to section, the way your eyes jump to bold text. It has labeled links, so instead of hearing "click here" twelve times I hear where each link actually goes. Its buttons say what they do. Its form fields say what they want. When I make a mistake, the error message is announced out loud instead of just turning a box red.
When a page is built badly, all of that disappears. I land on a site and the screen reader has nothing to hold on to. No headings. Links that say "read more." A row of icons that announce themselves as "unlabeled button." A menu that opens when a mouse hovers over it and never opens for a keyboard at all. I can sometimes still get through a page like that, but it is a bit like finding your way through an unfamiliar house by walking along the walls with your hand out. You get there. It takes a lot longer than it should, and you knock things over.
There is a standard for how this is supposed to work. It is called the Web Content Accessibility Guidelines, or WCAG, published by the World Wide Web Consortium. It has existed in one version or another for decades. The rules are not secret, and most of them are not hard. Label your images. Use real headings. Make everything work from a keyboard. Do not rely on color alone. The frustrating part of being blind online is not that accessibility is impossible. It is that it is usually possible, and people just did not do it.
I wrote about the bigger argument for this in Accessibility Is Not a Side Feature. This piece is more about the daily texture of it. What it actually sounds like.
The sound of a good day
I want to start with what works, because a lot does.
On a good day I can read the news, answer email, message my wife, check my bank, research a school assignment, upload a song, and post to social media without any help from anyone. That is not a small thing. I grew up at a time when a lot of that would have required a sighted person sitting next to me, and plenty of blind people older than me remember when it required a sighted person for almost everything.
Much of my own work runs through a screen reader. I make music as Mr. Blindbandit, and I do it without looking at waveforms, because I cannot. I navigate audio software through keyboard commands and spoken feedback, and I listen. I built and maintain my own website the same way. When people are surprised by that, I understand why, but it is also just how I work. I have written a little about it on the site under screen readers and music work and creating without sight.
A good day has a certain rhythm. The voice is fast and even. Pages behave the way I expect. I press H and land on a heading. I press Tab and move to the next control. I type, and the characters echo back to me. When something loads, I hear it. There is very little friction, and so I almost stop noticing the technology. That is the goal of accessibility, really. Not a special experience. An ordinary one.
The sound of a bad day
A bad day sounds different.
It sounds like a sign-up form where every field is announced as "edit text," with no label, so I do not know which box wants my name and which wants my password. It sounds like a checkout page that throws an error I cannot find, because the message appeared somewhere on the screen without being announced. It sounds like a pop-up that traps my focus, so that no matter how many times I press Tab, I keep circling the same three buttons and cannot get back to the page underneath. It sounds like an entire app built on a canvas that the screen reader cannot see into at all, so the voice just goes quiet.
Silence is one of the worst things a screen reader can do. It means there is something there and I am not being told what it is.
And then there are the images. So much of the internet is pictures now. Memes, screenshots of text, product photos, flyers for events, charts in the middle of articles, photos a friend sends with no caption. When an image has alt text, a short written description in the code, the screen reader reads it and I get the idea. When it does not, I hear "image," or worse, the file name, which is usually a string of numbers and letters that means nothing to anyone.
Screenshots of text are their own small cruelty. Someone writes three paragraphs, takes a picture of the paragraphs, and posts the picture. To most people it looks like text. To me it is a locked box.
None of these things are dramatic. Nobody is being cruel on purpose. That is kind of the point. Most of the barriers I run into were built by people who never pictured someone like me using what they made.
CAPTCHAs, or proving I am human
If you want to understand the specific absurdity of being blind online, spend some time with CAPTCHAs.
A CAPTCHA is the test a website gives you to prove you are not a bot. For years the most common version was a picture: squiggly letters to type, or a grid of photos where you are supposed to click every square with a traffic light or a bus or a crosswalk. By design, those tests depend on seeing. The whole premise is that a human can look at the picture and a machine cannot.
There are audio alternatives, and I use them. But audio CAPTCHAs have their own problems. The words are often distorted on purpose, buried in background noise so that software cannot pick them out, which also means a person listening carefully with good headphones sometimes cannot either. I will play one, then play it again, then ask for a new one. On a good site it works on the second or third try. On a bad site there is no audio option at all, and a page that was otherwise perfectly accessible slams shut at the last step.
Newer systems try to judge whether you are human in the background, based on how you move a mouse or how you behave on the page, so that most people never see a puzzle. That can help. It can also backfire on people like me, because I do not use a mouse, and a screen reader moving through a page in a straight line does not always look like what the system thinks a human looks like.
There is something that stays with me about being asked to prove my humanity by a test I physically cannot take. It is not a big wound. It is a small, repeated one. The internet asks, many times a week, whether I am real, and the way it asks assumes I can see.
When the machine started describing the world
The biggest change in the last few years has been AI, and I want to be careful here, because this is the part people either oversell or dismiss.
For most of my life, if I got a photo with no description, that was the end of it. I could ask someone what it was. I could guess from context. Or I could let it go. Now I can hand that photo to an AI model and ask what is in it, and within a few seconds a voice tells me. Apps like Be My Eyes, which has long connected blind people with sighted volunteers over video, and Microsoft's Seeing AI have added this kind of automatic description. General assistants like ChatGPT can do it too. I can point my phone's camera at a piece of mail, a box in the kitchen, a screen that is not talking, and get a description back.
I also use AI in a way people do not always expect: as a way to organize my own thoughts. I have used ChatGPT's voice mode as an accessibility tool, talking through my experiences out loud and getting help organizing and editing them into writing. I have said so openly on essays on this site, like A Journey Beyond Blindness. The experiences and opinions are mine. The tool helps me get them onto the page in a shape other people can read, the same way a screen reader helps me get a page into a shape I can hear.
That is real independence. I do not want to downplay it. There are small daily tasks that used to require asking someone, and now they do not. Asking for help is fine. I do it all the time, and independence has never meant refusing help. But there is a particular dignity in being able to find out what is on a piece of paper in your own hands without waiting for someone else to be free.
What AI still gets wrong
Here is the other side.
AI descriptions are guesses. Good guesses, often. But they are delivered in the same calm, confident voice whether they are right or wrong, and I usually have no way to check. If an AI tells me a document says one thing and it actually says another, I may not find out until it matters. For something like reading a menu, that is a small risk. For something like a medication label, a bill, or a legal form, it is not.
So I have learned to use it like a friend who is helpful but sometimes makes things up. I trust it for low-stakes things. For high-stakes things, I cross-check, or I ask a person I trust, or I find the text in a form a screen reader can read directly.
There is also a quieter risk that worries me more. Some companies may start treating AI as a replacement for building accessible things in the first place. Why write alt text if a model can describe the image? Why label the buttons if an AI can guess what they do? I understand the temptation. But a description written by the person who made the image will always be better than a machine's guess, because the person knows what they meant. And a guess layered on top of a broken page is still a broken page. AI can be a bridge over a gap. It should not become an excuse to stop filling the gaps.
The best version of the future, to me, is one where AI helps with the messy real world, the unlabeled photo, the paper letter, the sign on the wall, and the websites and apps are simply built correctly to begin with, so I do not need the bridge there at all.
Speed, patience, and the labor nobody sees
People who watch me use a phone for the first time are usually surprised by how fast the voice talks. I keep it at a speed that sounds like a blur to most listeners. That is not showing off. It is efficiency. When everything comes to you one line at a time, speed is how you keep up.
But speed only helps on a page that makes sense. On a page that does not, I slow down. I try a different navigation mode. I jump by form fields, then by links, then by headings, then line by line, looking for the one element that will let me move forward. Sometimes I switch browsers. Sometimes I switch devices, because a site that is broken in one place works in another. Sometimes I give up and come back later, or I find a different company that built its site better and give it my business instead.
That hidden work is the part sighted people almost never see. It is not dramatic. It is a thousand small detours. Over a day they add up. Over a year they add up to a lot of time I could have spent making music, studying, or being with my family.
And it affects how much energy I have left for everything else. When a site is accessible, I feel like any other customer. When it is not, I am a customer who also has to be a detective, a tester, and occasionally an advocate who writes to the company to explain what is broken. On the site, I put together a page on keyboard and screen-reader navigation partly because I know from experience how much it matters to be told, clearly, how a site is supposed to work.
What I would ask of the people who build things
I do not think the answer is complicated. It is mostly attention.
If you build websites or apps, try using your own product with a screen reader, even for ten minutes. Turn on VoiceOver or TalkBack on your phone, or install NVDA on a Windows machine, close your eyes, and try to sign up for your own service. You will learn more from that than from any checklist. Then use the checklist too. WCAG exists for a reason.
Write alt text that says what matters about an image, not just what is in it. Use real headings. Label every button and every field. Make sure everything works from a keyboard. If you use a CAPTCHA, make sure there is a way through it that does not depend on sight. If you post text, post it as text.
If you are not a developer, the list is shorter. Add a sentence describing the photo when you post it. Do not post screenshots of paragraphs when you can paste the paragraphs. If someone tells you something you made is not accessible, believe them. Most of us are not trying to start a fight. We are trying to use the thing you made.
The door and the room
I think about access a lot, not only as a user, but because I am trying to build something around it. The Mr. Blindbandit Access and Opportunity Foundation is still being built, still in formation, and a big part of the reason it exists is the gap I have been describing: the space between a technology existing and a person actually being able to use it.
Here is the way I think about it. Accessibility is the door. It lets you in. But the reason you wanted to get in was never the door. It was the room. The job application, the class, the song you wanted to release, the conversation with someone on the other side of the world. The internet has become the room where a huge part of life happens. For blind people, the door is sometimes wide open, sometimes stuck, and sometimes painted onto a wall.
In 2026, I can do more online by myself than any generation of blind people before me. I can hear my daughter's photos described. I can study, build, and publish from the other side of the world from where I grew up. I also still spend part of every week proving to a website that I am human, guessing at unlabeled buttons, and waiting for an AI to tell me what someone could have just written down.
Both of those things are true at once. The voice in my ear tells me all of it, one line at a time.
Follow Mr. Blindbandit
If this piece meant something to you, or if you build things and want to talk about making them work for more people, you can find me here:
More official links and business contact information are on the Connect page.