Week 3: ChatGPT

Learning objectives

By the end of this week, you will be able to:

1. Apply general screen reader exploration skills, including your screen reader's list of links and interactive elements (sometimes called an elements list), heading navigation, edit field navigation, button navigation, and Say All, to a new website, ChatGPT, with confidence rather than starting over from scratch.
2. Locate ChatGPT's message box, send control, and settings area independently, using exploration rather than memorized, possibly outdated instructions.
3. Explain in plain language what it means that ChatGPT offers different "model" options, and why a beginner does not need to worry much about choosing one at first.
4. Describe what custom instructions are and how they let you state a standing preference once instead of repeating it in every conversation.
5. Explain why voice mode is a genuine accessibility feature, not just a convenience, and describe at least two honest caveats to using it.
6. Upload a file or image to ChatGPT and ask a question about its contents, connecting this to image-description and optical character recognition (OCR) concepts.
7. Compare ChatGPT and Claude on at least three separate dimensions, such as file handling, memory approach, and voice mode, and explain how you would choose between them for a specific task.

Lesson

What ChatGPT is, in the vocabulary you already know

ChatGPT is an AI chat tool built by a company called OpenAI. Underneath its interface, it works on the same basic ideas that apply to AI chat tools in general: it is built on a large language model, or LLM, a system trained on huge amounts of text so that it can predict and generate human-like language one piece at a time. When you type a message to ChatGPT, that message is your prompt, the instruction or question you give the AI tool. ChatGPT breaks your prompt, and its own reply, into small chunks of text called tokens, and it can only work within a limited amount of text at once, called its context window, just like other AI chat tools.

In other words, ChatGPT and Claude are not two completely different kinds of technology. They are two different products, built by two different companies, both offering a chat-style front door to a large language model. Think of it the way you might think of two different email providers: the underlying idea, sending and receiving messages, is the same, but the interface, the exact menu names, and a handful of special features differ from one to the other. Learning ChatGPT is mostly a matter of applying what you already understand about how AI chat tools work in general, while paying attention to the specific features that make ChatGPT distinct.

ChatGPT is one of the most widely used AI chat tools in the world, which means many workplaces, coworkers, and job listings may reference it by name specifically, not just generically as "an AI tool." That makes comfort with ChatGPT valuable on its own, separate from your growing comfort with Claude. Employers do not usually care which single AI tool you know best. They care that you can sit down in front of an unfamiliar AI interface and figure out how to use it productively and independently, which is precisely the skill this week is building through direct practice, not through memorizing one specific product forever.

Applying your screen reader exploration skills to a new website

Before learning anything specific about ChatGPT, it helps to restate a principle that matters a great deal here: interfaces change. Menu names, button labels, and page layouts on any actively developed website, including ChatGPT, can shift between one visit and the next as the company updates its product. Rather than memorizing a fixed map of exactly where every control sits today, a map that could be wrong by the time you read this, your most reliable tool is a general set of exploration skills that work across almost any modern website.

When you open ChatGPT's website for the first time, or any time its layout feels unfamiliar, start with your screen reader's list of links and interactive elements, sometimes called an elements list. Most screen readers offer some version of this feature; check your own screen reader's shortcut list or documentation for the exact key combination that opens it. This feature brings up a list of the headings, links, form fields, or landmarks on the page, depending on which view you choose, letting you scan the whole page's structure in one place instead of guessing your way through it linearly. Use it to locate the message box, listed as a form field, and any nearby buttons, such as a send control or an attachment control.

Beyond the elements list, most screen readers also offer quick navigation commands for moving directly between headings, edit fields (places you can type text), buttons, and landmark regions, such as a main content area or a navigation sidebar, without opening a list first. The exact keys differ from one screen reader to another, so check your own screen reader's documentation or shortcut list for the current commands. Whatever the exact keys turn out to be, the underlying idea is the same everywhere: your screen reader's heading navigation lets you jump heading to heading, your screen reader's landmark or region navigation lets you jump between major page sections, and similar commands exist for edit fields and buttons.

Most AI chat websites, ChatGPT included, tend to follow a broadly similar pattern: a message box near the bottom of the main content area, a send control right next to it, a history or list of past conversations somewhere else on the page, often in a sidebar landmark, and a settings or account area reachable from a menu, often near the top of the page. Treat that description as a general, common pattern to expect, not a guaranteed, exact layout, since ChatGPT's specific arrangement can change over time. If something is not where this lesson suggests, trust your exploration tools over this lesson's wording; a five-minute pass with your screen reader's list of links and interactive elements will almost always find what you are looking for.

Starting a conversation and reading the reply

Once you have located the message box using edit field navigation or your screen reader's elements list, click into it, or activate it with Enter if focus lands on it as a non-editing control first, and type a short, clear message, exactly the way you would with any AI chat tool. Many AI chat websites let you send a message either by pressing Enter while your cursor is in the message box, or by moving to a nearby send button with Tab and activating it with Enter or Space. Try both approaches once so you know which one works reliably in the version of ChatGPT you are using, since this detail, like page layout, is exactly the kind of thing that can change and is worth confirming for yourself rather than assuming.

After you send a message, ChatGPT's reply usually appears below your message inside the same page, without a full page reload, similar to Claude and to many modern chat websites. Screen readers rely on something called a live region to announce this kind of change, a part of a webpage marked so that a screen reader notices and reads new content automatically without you needing to navigate to it yourself. In practice, live region support varies from one site to another and even from one browser to another, so do not assume ChatGPT will always speak its reply the instant it finishes appearing. If nothing seems to happen after sending a message, try moving to the bottom of the page, or use your screen reader's heading navigation to jump toward the newest reply, then read it using Say All, or your screen reader's continuous reading command. Check your own screen reader's shortcut list for the exact keys that move to the end of a page and start continuous reading.

Give yourself permission to pause and reread a reply, or to ask a follow-up question, the same way good conversation strategies encourage with any AI chat tool. A conversation with ChatGPT is not a one-shot exam; it is closer to a back-and-forth exchange, and asking it to shorten a reply, explain a term, or restate something in a different way is a completely normal and expected part of using the tool well, not a sign that you did something wrong the first time.

Understanding ChatGPT's model options, without memorizing specific names

Somewhere in ChatGPT's interface, usually near the top of the message area or inside a menu, you will likely find a control that lets you choose between different underlying models, sometimes described as different versions or engines behind the same chat tool. In plain language, a model in this sense is simply which particular version of the underlying AI system answers your message. Different models can differ in how fast they reply, how much they can handle in one context window, and how carefully they reason through a complex request, similar to choosing between a quick, casual assistant and a slower, more thorough one for a harder task.

This course deliberately does not name or describe specific ChatGPT models, because the exact lineup of models offered changes over time, sometimes more than once a year, and any specific name printed here could be retired or renamed by the time you read it. What matters far more than any specific model name is understanding the general idea: a model selector exists, it is worth knowing how to find using the exploration skills from earlier in this lesson, and as a beginner you generally do not need to agonize over which one to pick. The tool's default option is built to be a reasonable, general-purpose choice for most everyday tasks, and switching models is something you can explore later, once you are comfortable with the basics, rather than a decision you need to get right on day one.

If you do want to explore the model selector, use the same approach you would use for any unfamiliar control: locate it with your screen reader's elements list or by navigating buttons, activate it, and let your screen reader read the options it presents rather than relying on this lesson, or any other written source, to tell you exactly what you will hear. This habit, checking the live interface instead of trusting a fixed written description, is one of the most durable skills this whole course is trying to build in you, and it applies just as strongly here as it does anywhere else you rely on a live, changing interface, such as when researching job listings online.

Custom instructions: telling ChatGPT your preferences once

An idea worth returning to now is the concept of giving an AI tool reusable instructions, standing preferences you state once instead of repeating in every single conversation. ChatGPT offers a feature generally built around exactly this idea, usually called custom instructions, typically reachable from a settings or account menu near the top of the page. Custom instructions let you write a short statement of things you want ChatGPT to remember about how you like to work, so you are not retyping the same preferences at the start of every new chat.

For example, you might write custom instructions explaining that you are a screen reader user and prefer plain text responses organized with clear headings, that you would rather avoid tables unless you specifically ask for one, or that you want replies kept reasonably concise unless you ask for more detail. Once saved, ChatGPT is meant to apply these preferences automatically to new conversations, without you needing to explain your situation over again each time you open a fresh chat. This is a small feature with a large practical payoff for accessibility specifically, since screen reader users often benefit from output formatted differently than a sighted user might request by default.

You may meet this exact idea again in much greater depth later in this course, in a lesson that teaches you to build a complete personal instruction set, including saved files and reusable prompts, that works across multiple AI tools, not just ChatGPT's built-in custom instructions feature. Treat what you learn about custom instructions this week as your first hands-on encounter with a concept that becomes a serious productivity and accessibility tool later on. As with the model selector, find the current custom instructions area using your screen reader's elements list or heading navigation rather than memorized menu names, since account and settings menus are exactly the kind of interface detail that shifts over time.

Voice mode as a genuine accessibility win

Many AI chat tools, ChatGPT included, offer some form of voice mode, a way to speak your message out loud instead of typing it, and to hear the AI's reply spoken back to you instead of reading it as on-screen text. For a sighted user, voice mode is often marketed mainly as a convenience, something to use while driving or cooking. For a blind or low vision user, voice mode can be something more significant: a way to interact with an AI tool that does not depend on screen reader text navigation at all, for that particular exchange.

This matters because screen reader navigation, while a powerful and reliable skill you are building throughout this course, still takes real mental effort: locating a message box, typing a message, waiting for a reply, and reading that reply with Say All or by moving through headings. Voice mode can shortcut all of that for a quick, simple exchange. Asking a spoken question and hearing a spoken answer, with no typing and no navigating a page at all, can feel closer to a natural conversation and can reduce fatigue, especially on a phone, where typing is often slower than on a full keyboard. This is exactly the kind of feature this course means when it calls AI a genuine equalizer: a feature built for general convenience turns out to remove a specific, real barrier for a screen reader user.

Voice mode also comes with honest, practical caveats worth naming clearly rather than glossing over. Background noise, from a busy office, a public street, or a loud household, can make speech recognition less accurate, sometimes badly enough to be more frustrating than typing would have been. Speaking your questions and hearing replies out loud is not private in a shared space, such as an open office, a train, or a room with family members nearby, so voice mode is best used when you have a reasonably quiet moment and are comfortable with anyone nearby overhearing both sides of the exchange. And because exact voice mode controls, activation methods, and availability can differ between the ChatGPT website, its mobile app, and its desktop app, and can change over time, the reliable approach is the same one this lesson keeps returning to: explore the current interface using your screen reader's navigation tools to find out what is available to you right now, rather than assuming a specific voice mode feature works a specific way based only on this lesson.

File handling: uploading a document or image and asking about it

Two accessibility uses of AI are especially valuable for blind and low vision users: image description, having an AI tool describe the visual contents of a picture in words, and optical character recognition, or OCR, having an AI tool read printed or handwritten text out of an image or scanned document. ChatGPT, like other AI chat tools, generally supports uploading a file or image directly into a conversation and then asking questions about it, which puts both of these skills directly into your hands inside this specific tool.

The general pattern for finding an upload or attachment control is the same exploration approach used throughout this lesson: look near the message box for a button, often found using button navigation or your screen reader's elements list, that opens a standard file selection dialog when activated. That file dialog is a normal operating system control, the same kind you already use in other programs, so your existing screen reader skills for browsing folders and selecting a file transfer directly, with nothing ChatGPT-specific to learn there.

This feature is worth practicing deliberately, not just noting as a fact. A photo taken with a phone camera, a scanned letter, or a PDF menu can all become genuinely readable to you in seconds once you know how to attach a file and ask a clear question about it, turning ChatGPT into a practical, everyday accessibility tool well beyond its role as a conversational chat assistant. As with every feature in this lesson, exact file size limits, supported file types, and the attachment control's exact label can change over time, so treat any specific number or label you encounter while exploring as current information worth noting for yourself, not as a fixed fact to memorize permanently from this lesson.

ChatGPT and Claude side by side: choosing the right tool for the task

You now have real, hands-on experience with two major AI chat tools, Claude and ChatGPT, which puts you in a good position to compare them honestly rather than relying on secondhand opinions. This comparison is not about declaring one tool universally better than the other. It is about noticing genuine differences so you can make a deliberate choice for a specific task, the same practical, evaluative mindset worth bringing to judging any AI output.

On general conversational strength, both tools are capable, well-built AI chat assistants that can draft, explain, summarize, and answer questions in natural, readable language. Any specific claim about one being consistently "smarter" or "better at writing" than the other tends to age poorly, since both companies update their underlying models frequently. A more durable habit is to try the same prompt on both tools occasionally and notice, for yourself, which reply you found clearer, more useful, or better organized for your own purposes that day, rather than trusting a fixed ranking from any outside source, including this course.

On file and upload handling, both tools generally let you attach a document or image and ask questions about it, as this week's file handling section described. The exact file types supported, size limits, and number of files allowed at once can differ between the two tools and can change over time on either one, so if a particular file does not work in one tool, it is often worth simply trying the other before assuming the task cannot be done at all.

On memory and standing preferences, the two tools take related but distinct approaches. Claude offers memory features, which can automatically retain relevant facts about you and your past conversations over time as you use the tool. This week's lesson introduced ChatGPT's custom instructions, a standing statement you write once and save deliberately, rather than something built up automatically through use. Neither approach is strictly better; automatic memory can feel more effortless, while a written custom instructions statement can feel more predictable and easier to review and edit, since you can always reread exactly what you told the tool. Understanding both approaches conceptually, rather than only ever using one, prepares you well for building your own comprehensive, portable instruction set that works across tools later in this course.

On voice mode, ChatGPT is widely known for a strong, well-developed voice mode, and this lesson framed it as a genuine accessibility feature worth exploring deliberately. Whether Claude offers an equivalent voice feature, and how fully developed it is, is exactly the kind of detail that can change quickly, so the right move is to check the current state of both tools' apps yourself using the exploration skills from this lesson, rather than assuming permanently that only one of the two tools offers voice interaction.

Finally, on choosing a tool for a specific task, a genuinely practical habit is to think about the shape of what you need before you open either tool. A quick, spoken question you want answered while your hands are busy, such as a fast fact check or a short piece of advice, may be a strong candidate for voice mode, wherever it is currently strongest. A long, evolving written project you will return to repeatedly over days or weeks, such as a resume you are drafting across several sessions, may be a strong candidate for Claude's Projects feature, which is specifically built to hold files and context across many conversations in one organized place. Neither of these examples is a rule you must follow exactly; they are illustrations of the deeper habit worth building: naming what shape of task you are facing, then picking whichever tool's strengths, based on your own recent, hands-on experience, best fit that shape, rather than defaulting out of habit to whichever tool you happened to learn first.

Carry this comparison habit forward deliberately as you continue exploring other AI tools. The goal is not to memorize every feature of every tool, but to build a genuine, personally tested sense of which tool fits which kind of task, based on your own exploration rather than on any single source telling you the answer.

Key terms from this week

Use this list to review the vocabulary introduced in this lesson before you start the exercises and the test.

Custom instructions

A ChatGPT feature that lets you write a standing statement of your preferences once, so the tool applies them automatically in new conversations instead of you repeating them every time.

Voice mode

A feature that lets you speak your message out loud to an AI tool and hear its reply spoken back, instead of typing and reading on-screen text.

Model (in the sense of choosing an AI tool or model option)

A particular version of the underlying AI system answering your messages, which can differ in speed, thoroughness, and how much text it can handle at once.

Keyboard-only exercises

These exercises use only your keyboard. Where this lesson gives an example key combination, check your own screen reader's documentation to confirm the current command, since keystrokes can differ between screen readers. As always, if a specific menu name or button label in ChatGPT's current interface does not match this lesson's general description, trust your exploration tools, especially your screen reader's elements list and its heading, edit field, and button navigation, over any fixed wording written here.

Exercise 1: Locate the message box and have a short conversation

1. Open your web browser and go to ChatGPT's website, signing in if needed.
2. Open your screen reader's list of links and interactive elements (sometimes called an elements list), and switch to the form fields or form controls view to locate the message box. Check your screen reader's documentation for the exact command that opens this feature.
3. Close the elements list, move focus into the message box, and type a short, clear question, such as asking ChatGPT to explain an unfamiliar term in one sentence.
4. Send the message, either by pressing Enter in the message box or by finding a nearby send button using button navigation and activating it.
5. Read the reply using Say All, or your screen reader's continuous reading command, moving to the bottom of the page first if the reply does not announce itself automatically.

Exercise 2: Explore the custom instructions or settings area

1. From ChatGPT's main page, use your screen reader's landmark or region navigation, or your screen reader's elements list, to locate a navigation or menu region near the top of the page.
2. Explore that region with heading and button navigation to find a settings or account menu.
3. Open the settings menu and look for an area related to custom instructions or personalization, reading each option with Say All as you go.
4. If you locate the custom instructions area, read what it asks for, and, if you feel ready, draft one or two sentences describing your own preferences, such as requesting plain text formatted with clear headings.
5. Close the settings area without saving if you are only exploring, or save your draft instructions if you are ready to keep them.

Exercise 3: Try voice mode, or upload a file and ask a question about it

1. Choose one of the two options below, based on what is available to you and what you would like to practice.
2. Option A, voice mode: in a quiet, private moment, use your screen reader's elements list or button navigation to locate a voice mode control, activate it, and ask a short spoken question, then listen to the spoken reply.
3. Option B, file handling: use button navigation to locate an attachment or upload control near the message box, activate it, and select a document or image from your computer using the standard file dialog.
4. If you chose Option B, type a question about the file in the message box, such as asking for a description of an image or a summary of a document, and send it.
5. Read the reply with Say All, and write one sentence for yourself noting whether the feature you tried felt useful, and why.

Portfolio project: Claude versus ChatGPT, a hands-on comparison

This week's portfolio piece is a short written comparison of the two AI tools you have now used hands-on: Claude and ChatGPT. Base it on your own experience from this week's exercises, not on outside opinions.

1. Open Notepad or your word processor of choice.
2. Write a heading at the top: "Claude versus ChatGPT: My Comparison."
3. Write one paragraph describing your experience navigating ChatGPT's web interface, compared to your experience with Claude.
4. Write one paragraph comparing the two tools on at least three specific dimensions, such as file handling, memory or custom instructions, and voice mode.
5. Choose two or three specific real tasks you might need AI help with, such as a quick voice question, a long written project, or describing a photo, and note which tool you would choose for each one, and why.
6. Save the file with the name claude-versus-chatgpt, in your portfolio folder.

Weekly test

Answer key

Each answer below includes a one-sentence explanation drawn from this week's lesson.