Frequently Asked Questions

If you are new to website building, some of the terms about AI may be unfamiliar. These questions bring together topics covered in software makers' FAQs and accessibility documentation, along with questions that help explain my own website. The answers are written in my own words. The tools listed are examples people can explore; they are not a list of products I personally use or a promise that every feature will work for everyone.

Understanding AI coding

1. What am I doing when I build a website with AI?

I describe what I want the website to say and do, supply my personal text, and ask AI to help produce the code. I can then ask for changes, explanations, and help resolving problems. For example, I asked for a homepage, an About Me page, and links between them. The AI helps carry out those requests, while I choose the content and direction.

2. Is vibe coding the same as all AI-assisted coding?

The terms overlap, but they are not always used identically. Vibe coding often emphasizes describing a desired result and refining what AI produces, sometimes with little engagement with the code itself. AI-assisted coding is a broader description that also includes carefully reading, editing, and learning from generated code. My goal is to use the assistance as a guide while developing my understanding. Research on programming through conversations with AI describes a process involving prompts, evaluation, and revisions.

3. Does AI write every word and make every decision for me?

No. The facts about my life come from me. I decide what information belongs here and what I want the site to become. AI can help organize my writing, improve the wording, and create code, but I can ask for changes when the result does not reflect what I mean. That is why my About Me and Vibe Coding pages explain the assistance openly.

4. What is a prompt, and what makes one useful?

A prompt is the instruction or question you give an AI tool. A useful request explains the goal, provides the relevant content, and names important requirements. For example: “Use this biography to create an About Me page. Add a clear title, a homepage link, and headings that a screen reader can navigate.” If the response misses something, follow up with the specific correction rather than assuming the tool understood everything.

5. Is the AI a person, and does it understand my website perfectly?

An AI coding assistant is software. It generates responses and code from the information available to it. A convincing explanation can still be wrong, and an assistant may misunderstand what you intended. Give it the relevant files or details, ask questions when something is unclear, and check the result. Its confidence is not evidence that a page works.

Tools and operating systems

6. What kinds of tools can people use for AI coding?

There are several categories. A conversational assistant can explain a code example or help plan a page. An editor lets you open and change the actual files. An AI coding assistant integrated into an editor can suggest changes in the project. A coding agent may also edit files or run development commands when permitted. The right combination depends on the task and on how comfortably you can operate its interface.

Examples include GitHub Copilot, which offers coding assistance in development workflows, and Anthropic's Claude Code, an AI coding tool. I have used OpenAI's Codex to help create and update this website. These are examples, rather than a ranking of which product is best.

7. What is Visual Studio Code, and is it an AI tool by itself?

Visual Studio Code, often shortened to VS Code, is a code editor available for Windows, macOS, and Linux. Its core editing features can be used without signing in. AI capabilities, such as GitHub Copilot, add another layer of assistance and may have their own account requirements and usage limits. An editor and an AI service are related parts of a workflow, but they are not the same thing. See the official VS Code FAQ.

8. Can sighted, blind, and low-vision people use the same editor?

Yes, people can often work in the same editor with different ways of accessing it. A sighted person may use the visual interface; a blind person may use a keyboard and screen reader; someone with low vision may enlarge the interface or choose a high-contrast theme. VS Code documents keyboard navigation, screen reader support, zoom, and high-contrast options. That does not guarantee that every extension or AI panel is equally accessible. See VS Code's accessibility guide.

9. Which operating systems can people use?

Windows, macOS, and Linux are common desktop choices for website development. The operating system manages the computer and runs applications; it is separate from the website's code. You do not have to use the same operating system as your visitors. A webpage hosted on a server can be opened in browsers on many kinds of devices.

Some people also use ChromeOS on Chromebooks or access browser-based tools on iOS, iPadOS, or Android. The available editors, development commands, and accessibility features differ. Mobile devices can be useful for reading, asking questions, and checking pages, while handling several project files and uploading them may be easier on a desktop or laptop. Check the requirements of the particular coding tool before choosing a setup.

10. Do these tools cost money?

Costs depend on the product and plan. Some editors and screen readers are free, while AI services may offer limited free access and paid subscriptions. Usage allowances and included features can change, so check the provider's current information. For example, GitHub's Copilot plans page explains its available plans. NVDA is available free of charge from NV Access.

11. Do I need internet access all the time?

You can edit a saved HTML file and open a self-contained page on your own computer without an internet connection. Uploading it to a hosting server requires a connection. Many AI coding services also rely on online processing, so offline editing and offline AI assistance are different things. Check the tool's documentation rather than assuming its AI features will work without internet access.

Coding without sight

12. What is a screen reader?

A screen reader is assistive software that presents information from an application through speech and, with supported equipment, a refreshable braille display. It can announce text, links, headings, buttons, and other controls. It also helps a person navigate the interface. It is much more than a recording that reads a whole screen from top to bottom. GNOME's description of Orca explains speech and braille access to a graphical desktop.

13. Which screen readers are available on Windows?

NVDA, from NV Access, is a free screen reader for supported Windows PCs. JAWS, from Freedom Scientific, provides access to Windows and many applications. Narrator is Microsoft's screen reader built into Windows. They have different commands and capabilities; check current system requirements and licensing before choosing one.

14. What can people use on Apple devices, Linux, Android, or a Chromebook?

VoiceOver is built into the Mac, and Apple also provides VoiceOver on iPhone and iPad. Orca is a free, open-source screen reader associated with the GNOME desktop and used on compatible Linux setups. TalkBack is Google's screen reader included on Android devices. Chromebooks include ChromeVox. Support depends on the operating system, application, and particular task; these are not interchangeable programs.

15. How can a blind person read and edit code?

Code is text, so a screen reader can provide access to it in a compatible editor. A person can move through characters, words, and lines using keyboard commands. Punctuation matters in programming, and screen reader settings can affect how much punctuation is spoken. Indentation and editor announcements also help communicate structure. A supported braille display can provide another way to examine the text. The skill involves learning both the editor and the access tools.

16. Is a screen reader the same as AI, dictation, or a magnifier?

No. A screen reader helps you access information and controls. An AI assistant can generate code or explanations. Dictation turns spoken words into input. A magnifier enlarges content for someone who uses vision. A person might combine these tools, but each has a different job. For example, Apple's accessibility documentation describes VoiceOver, Zoom, and Voice Control as separate features. See Apple's Mac accessibility overview.

17. How can someone check whether an AI coding interface is accessible?

Try the main tasks with the setup you actually use: reach the prompt field, send a request, read the answer, inspect proposed changes, and move back to the editor. Check whether focus is clear, controls have understandable names, and keyboard navigation lets you leave each panel. VS Code provides an Accessible View for supported content and accessibility help for different parts of its interface. Consult its accessibility documentation for current commands. A feature working for one screen reader does not establish that every workflow works for everyone.

18. Can AI make a website accessible automatically?

AI can help write headings, labels, and keyboard-friendly controls when you ask for them. Those changes still need to be checked. A useful test is to navigate the page without a mouse, follow its links, and examine its structure with a screen reader. Also consider text enlargement, readable contrast, and understandable instructions. Accessibility is something to evaluate in the finished experience, rather than a label to accept because a tool used the word “accessible.”

Learning and checking the work

19. Do beginners need to know how to code before starting?

You can begin with a small goal and ask for explanations along the way. For a simple page, learn what a heading, paragraph, link, and file path do before adding complicated features. Ask the assistant to explain a change in plain language and show where it belongs. Starting is possible without knowing every detail, but gaining understanding helps you recognize mistakes and maintain what you build.

20. Is AI-generated code guaranteed to be correct or safe?

No. GitHub's guidance describes limitations in generated suggestions and the need to review and validate them. Code can contain errors or security problems even when it looks plausible. The amount of checking should match the feature: a biography page is different from a system handling logins or payments. See GitHub's responsible-use guidance for coding suggestions.

21. What should I share with an AI assistant, and what should I leave out?

Provide the content needed for the task, a clear description of the problem, and relevant code where appropriate. Avoid putting passwords, private keys, or other people's confidential information into a prompt. Products have different data policies and settings, so read the documentation for the service you use. For Claude Code, its official user FAQ links to guidance about permissions and data usage. Do not assume every AI service handles information the same way.

22. Does creating a page mean it is already online?

No. A page can exist only as a file on your computer. To publish it, the file must be uploaded or deployed to the server that hosts the website. A domain such as dougoliver.net is the address visitors use to reach that site. For my current pages, index.html is the homepage and about/index.html is the About Me page. The links must point to the correct locations after upload.

23. Why might visitors see an older page after an update?

A browser or server can keep a cached copy to load a page more quickly. An uploaded update may therefore take time to appear at the usual address. Checking the server file and checking the public webpage are separate steps. A fresh request or a link with a version parameter can help determine whether the new page is available, but the ordinary visitor address also needs to be checked.

24. What would be a sensible first project for someone interested in this?

Choose one small page with a clear purpose: a personal introduction, a reading list, or a description of a hobby. Write the content first, ask for a simple structure, and check the headings and links. Learn how to save the file and open it in a browser before adding more pages. If you use a screen reader, include keyboard and screen reader checks from the beginning. You can grow the project as you become more comfortable with the tools.

25. What do I want visitors to understand about my work?

I am using technology to put my own ideas and experiences online. AI assists with writing and explaining code, and I guide the content and direction. Being blind does not prevent someone from working with website files, but access to the tools matters. This website is also a learning process: I can ask questions, improve the wording, add pages, and develop a better understanding as it grows.

Costs, usage limits, and planning a budget

26. What expenses can come up when building a website with AI?

Several separate costs can be involved: an AI subscription, additional AI usage, a domain name, web hosting, and any paid tools or services the website needs. There may also be costs for a computer, internet access, assistive software, or a braille display, depending on what someone already has. These costs do not all apply to every project. A simple HTML website may need far fewer services than a site with accounts, a database, or a live radio stream.

The examples here explain possible expenses; they are not a statement of my personal bills. Before paying, check what is included, how renewal works, whether tax applies, and whether extra usage can be charged separately.

27. What are examples of AI subscription prices?

As checked in September 2026, GitHub lists Copilot Free at no subscription cost, Copilot Pro at US$10 per month, Pro+ at US$39 per month, and Max at US$100 per month. Paid plans have different allowances and features; a monthly price does not mean every kind of AI activity is unlimited. These are examples from one provider, not prices for every AI tool or a recommendation to buy the most expensive plan. See GitHub's current plan comparison and confirm the checkout total before subscribing.

28. What is a usage limit, and why can a paid plan still have one?

A usage limit controls how much of a service you can use during a particular period. Depending on the provider, limits can involve messages, processing allowances, credits, requests, or access to particular models. Some reset after a shorter session window; others apply weekly or monthly. A paid plan can provide a larger allowance while still having limits. You need to look at both the subscription price and the allowance that comes with it.

Long conversations, large amounts of project context, and demanding coding tasks can affect consumption. Anthropic explains that usage varies with factors such as conversation length and the features used, and that activity across its Claude product surfaces shares usage limits. See Claude's usage and length limits explanation. A limit on how much information fits into one conversation is also different from an allowance for how much work you can do over time.

29. What happens when the included allowance runs out?

What happens depends on the provider, plan, and billing settings. You may need to wait for a reset, use a different available model, upgrade, or purchase additional usage where offered. Some products let you continue through separately billed credits or extra usage. Check whether that option is enabled and whether a spending limit is available before starting a long task. Do not assume that reaching a limit will always stop charges automatically.

OpenAI's official coding-service pricing guidance explains included usage and options for additional credits. Use the account's current usage display for the applicable reset time and allowance rather than relying on someone else's experience.

30. What are tokens, credits, and API charges?

Tokens are units used to process text and other model inputs or outputs; a token is not necessarily one word. An API lets software call an AI service directly. API billing can charge for input and output usage at rates that vary by model. A subscription to a chat or coding product should not be assumed to include separate API usage. See OpenAI's API pricing documentation for its model-specific charges.

Credits are a provider's way of accounting for an allowance or purchased usage. One provider's credit is not automatically equivalent to another's. A small instruction can also lead a coding agent to make several model calls, inspect files, and revise its work, so “one prompt” is not always a predictable unit of cost. Check the billing details of the product you actually use.

31. What website and accessibility expenses should someone consider separately?

A domain is the site's address and normally has a renewal charge. Hosting provides the server space or platform that serves the pages. Introductory hosting offers may differ from renewal prices. Storage, bandwidth, email, databases, and other services may have separate allowances or fees. A radio station can also involve streaming hosting and music licensing expenses that are separate from creating its webpage; those depend on the station's circumstances.

Accessibility costs also vary. NVDA is available free from NV Access, and Windows includes Narrator. Other assistive software, hardware, training, or support may involve payment. A person does not need to buy every tool named on this page. Consider the equipment and access methods that fit the tasks you want to do.

32. How can someone plan their spending and use their allowance thoughtfully?

Make a short list of the services needed and separate recurring subscriptions from usage-based charges. For an illustrative budget, a US$10 AI subscription plus hypothetical US$8 monthly hosting and a hypothetical US$24 yearly domain would average US$20 per month before tax or extra usage. The hosting and domain figures in that calculation are made-up examples, not offers or my actual expenses. A yearly domain payment may still be due in one lump sum.

For the work itself, provide clear requirements and the relevant files, make focused changes, and check the result before requesting more. This can reduce unnecessary revisions, although it cannot guarantee a particular saving. Watch the usage display, review billing alerts, and use spending controls where the service provides them. Start with the allowance you have and increase spending only when you understand what additional access you need.

Tool and accessibility information checked in September 2026. Questions are adapted from topics in the linked official FAQs and documentation; they are not copied quotations or a ranked list of the most commonly asked questions.

Do any of these services offer unlimited usage?

Some offer unlimited use of specific features, but that does not mean unlimited access to every AI model or unlimited coding-agent work for one fixed price. The plans reviewed here do not establish an all-features, all-models unlimited AI coding subscription. Choosing a higher tier usually increases an allowance or adds features; check exactly what is included before paying.

Which services advertise unlimited features?

• GitHub Copilot: paid plans include unlimited code completions and next-edit suggestions. Other AI activity uses the plan's AI credit allowance, so this is not unlimited agent or chat usage. Check Copilot's plan details.
• Devin / Windsurf: the pricing page lists unlimited Tab completions and inline edits, including on Free. Agent work has quotas; paid plans can purchase extra usage. Check Devin's pricing.
• Zed: Pro includes unlimited edit predictions, but hosted AI model usage includes only a specified token credit allowance before extra charges. Its free editor permits use of your own API keys or external agents; their charges and limits still apply. Check Zed's pricing.
• Tabnine: its pricing describes unlimited usage when you supply your own on-premises model or cloud model endpoint. That does not make the model or infrastructure free. Tabnine-provided model access has separate token costs. Check Tabnine's pricing conditions.

Does “no daily limit” mean unlimited usage?

No. For example, Bolt Pro has no daily token limit, but its entry plan still has a monthly token allowance. Removing a daily cap does not remove the monthly budget. Similarly, unlimited projects, workspaces, or collaborators says nothing about the amount of AI processing included. See Bolt's daily and monthly allowances.

Can I keep working after my allowance runs out?

Some services offer extra credits or pay-as-you-go usage. That can let you continue, but it is additional paid usage, not an unlimited subscription. Cursor, for example, includes a set amount of model usage and bills on-demand usage separately after that amount is consumed. Other products may require waiting for a reset or choosing a different plan. See Cursor's usage-based billing explanation.

What should I check before choosing an “unlimited” plan?

Confirm which feature and models the word applies to, whether there are rate limits or acceptable-use restrictions, whether extra charges are automatic, and whether a spending cap is available. Check the provider's current terms and your account's usage screen. An expensive or enterprise plan is not proof of unlimited use.

Unlimited-feature information checked September 18, 2026. Read the subscription and usage disclaimer and service pricing comparison before choosing a plan.

AI coding services: plans and pricing

Pricing checked September 18, 2026. Prices are in US dollars before applicable taxes. Monthly prices mean monthly billing unless an annual commitment is explicitly stated. Promotions, limits, and plans can change; use each official source to confirm the current offer.

This is a broad guide to major coding assistants, coding agents, and AI app builders, not an exhaustive directory of every AI product. General chatbots can also write code. A free editor or free account does not necessarily include free model usage, hosting, or a domain.

How do the pricing structures differ?

• Subscriptions: a recurring payment includes a defined allowance.
• Per-user plans: each team member or assigned seat adds a charge.
• Credits and tokens: the amount of work consumes an allowance; models and complex tasks can cost different amounts.
• Pay as you go: additional model calls, cloud resources, or app traffic create usage charges.
• Annual contracts: a lower advertised monthly equivalent may require a full year of payment or commitment.
• Enterprise quotes: the price depends on a negotiated contract.

What about other tools and model providers?

New products and pricing changes make a complete list impossible to keep fixed. Continue's current website reports that it has been acquired by Cursor; check Continue's official update before relying on older standalone prices. Tools connected to OpenAI, Anthropic, Google, DeepSeek, Mistral, or other model providers may charge by model usage rather than a flat coding subscription. Use the provider's current rate card and account spending controls.

How should I compare the total cost?

Compare the subscription, included allowance, additional usage rates, billing period, and any hosting or domain charges together. Check that the interface works with your keyboard and screen reader before committing. A higher price does not guarantee a better fit or better accessibility.