foundations · Level 1
Using AI as an accessibility tool: reading, writing and seeing
Screen readers, dictation, live captions and image description all use AI now. A practical tour of what your device can already do, and how to check an AI description before you trust it.
Your next step
Start this course online
Register your start, then work through the lessons and exercises at your own pace. You can download the PDF separately. Where a scored exam is available, you will find it after the lessons.
Register or resume your course
Start with the essentials
The short answer
Most phones and computers already include AI accessibility tools built in and free: a screen reader that reads a page aloud, dictation that types what you say, live captions for audio, and image description for photos a screen reader cannot otherwise interpret. Each is genuinely useful, and each has a specific, learnable failure mode worth checking for before you rely on it.
What you will learn
- You will be able to turn on and use your device's built-in screen reader well enough to navigate a simple page or document.
- You will be able to use dictation to write text by speaking, and know the specific kinds of errors to check for afterward.
- You will be able to turn on live captions for a video call or a video, and know what situations make captions less reliable.
- You will be able to get an AI description of a photo and know what kind of detail it commonly gets wrong.
- You will have a short checklist for when an AI accessibility feature's output needs a human check before you rely on it for something that matters.
Who it is for
Anyone who wants to use, or wants to help someone else use, the AI accessibility features already built into their computer or phone. No technical background needed.
Before you start
- None. It helps to have read What is AI? if you are new to AI tools generally.
Learn in your browser · Chapter 01 of 05
What is already built in, and why it is AI now
Accessibility features used to be simple and rule-based. Most of the useful ones now use AI underneath, which is why they have become far more capable, and why they can also fail in AI-shaped ways.
Screen readers, dictation, live captions and image description are not new. What has changed is the technology underneath them. A screen reader used to read exactly what was marked up in a document's structure, nothing more. Modern versions use AI to make a reasonable guess at what an unlabelled button or image is for. Dictation used to require you to speak in a stilted, deliberate way to a system trained on a narrow vocabulary; modern dictation understands natural speech far better because it uses the same kind of AI model behind general-purpose chat tools.
This matters because it changes the failure mode. A purely rule-based tool fails obviously: it simply cannot do the thing. An AI-based tool often fails quietly: it produces a plausible-sounding answer that happens to be wrong, a caption that sounds right but named the wrong speaker, or an image description that confidently describes something that is not actually in the photo.
Learn in your browser · Chapter 02 of 05
Screen readers: hearing a page instead of reading it
Every major operating system has a screen reader built in, free, no extra software needed.
Windows includes Narrator, macOS and iOS include VoiceOver, and Android includes TalkBack. All three read the text on screen aloud, describe buttons and images, and let you navigate a page or app using keyboard shortcuts or gestures instead of a mouse. None of them require installing anything extra or paying for anything.
Try it yourself · Activity 01
15 minTurn on your screen reader and navigate one page
Use whichever device you have to hand.
- Turn on your device's built-in screen reader (search your system settings for 'screen reader', 'Narrator', 'VoiceOver' or 'TalkBack' if you are not sure where the toggle is).
- Without looking at the screen, use the screen reader's navigation commands to move through a simple webpage, heading by heading.
- Find one image on the page and listen to how the screen reader describes it. Note whether the description is genuinely useful or just says 'image'.
- Turn the screen reader off and compare: did you get a real sense of the page's structure, or did some of it clearly get lost?
Where did the screen reader struggle most: badly labelled buttons, unlabelled images, or a confusing page layout? That gap is usually the website's fault, not the screen reader's, worth knowing if you ever build a page yourself.
Learn in your browser · Chapter 03 of 05
Dictation: speaking instead of typing
Dictation has gone from a niche, error-prone tool to something genuinely reliable for ordinary writing, with one recurring weak spot.
Built-in dictation (Windows Voice Access, macOS Dictation, Android and iOS voice typing) now understands natural speech well enough for everyday writing: emails, messages, notes. You speak, and it types. It works offline on some devices and online on others, and it is free and built in on all the major platforms.
Try it yourself · Activity 02
15 minDictate a short paragraph and check it
Turn on dictation on your device.
- Dictate a short paragraph, three or four sentences, about something ordinary: your plans for the day, a description of the room you are in.
- Read back what was typed without correcting it first.
- Identify every word that was wrong. Pay particular attention to names, numbers and less common words.
- Correct the errors and note which kind of word caused the most mistakes.
Names, numbers and uncommon words are the classic weak spot for dictation. If you regularly dictate something with a lot of names or numbers, that is exactly where you need to read back carefully every time, not just for important documents.
Learn in your browser · Chapter 04 of 05
Live captions: reading what you cannot hear
Live captioning works well for clear, single-speaker audio and gets noticeably less reliable the further a real conversation drifts from that.
Live captions, built into video calling apps, many phones and video players, transcribe speech to text in real time using AI. They are genuinely useful for following a video, a call, or a meeting without relying on hearing every word clearly. They are not a word-for-word court transcript.
Scroll sideways to see every column.
| Situation | How reliable captions tend to be | What to watch for |
|---|---|---|
| One clear speaker, quiet room | Generally reliable | Occasional wrong word, rarely changes the meaning |
| Several people talking over each other | Noticeably worse | Captions can merge or misattribute who said what |
| Strong accent, unfamiliar name, technical term | Weakest point | Names and jargon are the most commonly mis-transcribed |
| Background noise, poor microphone | Degrades quickly | Whole phrases can be dropped or garbled |
Try it yourself · Activity 03
10 minTest captions against something you already know
Find a video or recording where you already know roughly what is said.
- Turn on live captions and play a short section, at least one minute.
- Compare the captions against what you know was actually said.
- Note specifically what kind of error appeared, if any: a wrong word, a missed word, a misattributed speaker.
- If a name or technical term appears, check whether the caption got it right.
If you were relying on captions alone for something important, a job interview, a medical appointment, would the errors you found have mattered? That is the honest test of whether captions alone are enough for a given situation.
Learn in your browser · Chapter 05 of 05
Image description, and a before-you-trust-it checklist
AI image description can turn a photo into a spoken or written summary, genuinely useful and worth knowing its specific blind spots.
Several tools, built into some screen readers and available as standalone apps, can describe a photo aloud: what is in it, roughly how it is arranged, sometimes readable text within the image. This is genuinely useful for identifying an unlabelled photo, reading a sign in a picture, or getting a general sense of an image a screen reader would otherwise just skip.
Try it yourself · Activity 04
10 minTest an image description against a photo you know
Use a photo you already know well, one with people, text or a specific detail in it.
- Get an AI description of the photo through your device's screen reader or a dedicated image description tool.
- Compare the description against what you know is actually in the photo.
- Note anything the description got wrong, invented, or missed entirely, especially any text in the image.
- If the image contains a sign, label or handwriting, check whether the described text matches exactly.
Text within an image (a sign, a label, a screenshot of a message) is one of the most error-prone things for image description to get exactly right. Treat any text an AI reads from a photo as a draft to verify, not a confirmed transcript, particularly for anything with numbers.
- Never treat an AI image description as a confirmed transcript of text within the photo; verify any numbers, names or instructions it claims to have read.
- Never assume an AI caption or description has covered everything in a photo or a recording; it summarises, and a summary can omit something that matters.
- For anything consequential, a document you are about to sign, a medical instruction, a safety notice, get a second confirmation rather than relying on one AI description alone.
- If a description or caption does not match what you expect, ask again or try a different tool rather than assuming your own understanding must be wrong.
Keep learning
The complete workbook
This workbook is a practical tour of the AI accessibility features already built into most computers and phones: turning on and using a screen reader, dictating text instead of typing it, reading live captions on a video call or video, and getting an AI description of a photo when you cannot see it clearly yourself. Each section covers how to turn the feature on, what it does well, and a specific way it can get things wrong, so you know when to double check rather than take the output on trust.
- 01What is already built in, and why it is AI nowRead here · Reading
Accessibility features used to be simple and rule-based. Most of the useful ones now use AI underneath, which is why they have become far more capable, and why they can also fail in AI-shaped ways.
- 02Screen readers: hearing a page instead of reading itRead here · 1 exercise
Every major operating system has a screen reader built in, free, no extra software needed.
- 03Dictation: speaking instead of typingRead here · 1 exercise
Dictation has gone from a niche, error-prone tool to something genuinely reliable for ordinary writing, with one recurring weak spot.
- 04Live captions: reading what you cannot hearRead here · 1 exercise
Live captioning works well for clear, single-speaker audio and gets noticeably less reliable the further a real conversation drifts from that.
- 05Image description, and a before-you-trust-it checklistRead here · 1 exercise
AI image description can turn a photo into a spoken or written summary, genuinely useful and worth knowing its specific blind spots.
Also inside: a 8-point checklist, a glossary of 8 terms and 8 questions and answers to test yourself. 4 hands-on exercises, each with a worked answer at the back where the workbook gives one.
The PDF is free. We ask for your name and email before downloading. Read the privacy notice at the download form. Free to download and use for personal learning, study groups and inside your own team. Please do not resell the workbooks or republish them as your own. Link people to trust-agent.ai instead.
Get your free PDF workbook
Enter your name and email, then download the complete workbook. This form also works without JavaScript.
Test yourself
Questions and answers
Do I need to buy anything to use these accessibility features?
No. Screen readers, dictation, live captions and image description are already built into every major phone and computer operating system, switched on in the accessibility settings, at no extra cost.
Is AI-generated live captioning accurate enough for a legal or medical record?
No. Live captions are built for following along in the moment, not for an accurate transcript. Overlapping speakers, accents and unfamiliar names make them noticeably less reliable, so a record that matters needs a proper transcription service, not live captions.
Why does dictation get names wrong so often?
Dictation predicts likely words from sound, and an unusual name has no common pattern to match against. It is a specific, predictable weak spot, which is why a careful read-back of names and numbers matters every time.
Can I trust an AI's description of a photo without checking it?
Not for anything consequential. Image description can miss or misstate details a sighted person would catch immediately, so treat it as a useful first pass to confirm, not a verified fact on its own.
Are these tools only for people with a permanent disability?
No. A screen reader helps someone with tired eyes at the end of the day, live captions help in a loud room, and dictation helps an injured hand. They are general-purpose tools that also happen to be essential accessibility features.
What is the single biggest risk across all four tools?
Treating a confident-sounding AI output as automatically correct. Every tool in this workbook fails in a specific, learnable way, and the fix is always the same: check before you rely on it for something that matters.
Does turning on a screen reader change how my whole device works?
It changes navigation (you move by swipe or keyboard commands and hear what is on screen) rather than replacing the apps themselves. Most people can turn it on, try it for one task, and turn it off again.
Why do live captions lag behind real speech?
The system needs a short window of audio to predict words accurately, so there is always a small delay. That delay grows when speech is fast, overlapping or unclear, which is when captions are also most likely to be wrong.
Write as you go
My notes
Use this scratchpad for your own notes, working or answers as you read. Notes are saved in this browser on this device. Downloading makes a local copy; it does not send your notes to Mickai.
Consolidate your learning
Review your work
Work through the exercises and questions, then compare your reasoning with the worked answers. A scored exam and verifiable certificate are not yet available for this course.
Review the questionsLearn the language
Key terms
- Screen reader
- Software that reads the text and describes the content of a screen aloud, letting someone navigate a device without needing to see it. Built into every major operating system.
- Narrator, VoiceOver, TalkBack
- The built-in screen readers on Windows, Apple devices and Android respectively, all free and included with the operating system.
- Dictation
- Software that converts spoken words into typed text in real time, now commonly powered by the same kind of AI models used in general-purpose chat tools.
- Live captions
- Real-time, AI-generated text transcription of spoken audio, shown on screen during a call, video or recording as it plays.
- Image description
- An AI-generated spoken or written summary of what a photo contains, used by screen readers and standalone apps to make images accessible to someone who cannot see them clearly.
- Misattribution
- When a caption or transcript assigns a piece of speech to the wrong speaker, a common error in live captions when several people talk over each other.
6 of the workbook's 8 terms. The complete glossary is in the workbook.
Follow the evidence
Sources and checks
Facts last checked: .
These workbooks use AI assistance. See how the workbooks are made.
Created by Mickarle Wagstaff-Irons - Micky Irons with the Mickai team. Published by Mickai LTD. Last updated 29 September 2026.
NextKeep going
Where to go next
Recommended for you
What is AI? A plain-English starter
A jargon-free introduction to modern AI: how chatbots work, what they are good and bad at, and how to use them safely from day one.
Recommended for you
Your first prompts: how to ask AI for what you actually want
Learn to ask AI for what you actually want: the six parts of a good prompt, how to improve a reply, and how to build a personal prompt library. Free tools only.