How to Make a Quiz with AI (And Share It as a Link)

Making a quiz with AI takes about five minutes. ChatGPT and Claude both generate a complete, interactive quiz as a single HTML file — questions, answer choices, score logic, and results screen all in one document that runs in any browser. With 900 million people using ChatGPT weekly, AI quiz generation is now a common task. The part that stumps people is what comes after: you have a quiz on your laptop, but your students or participants need a link.

Every top result for “how to make a quiz with AI” is either a quiz SaaS product page or a prompt guide that stops at “here is the HTML code.” None of them explain how to go from that HTML file to a URL that anyone can open — and none address the update problem: what happens when you need to fix a question without sending everyone a new link. This guide covers both.

What kind of quiz can AI actually build?

AI tools like ChatGPT and Claude produce self-contained interactive quizzes that run entirely in the browser. The output is a single HTML file combining markup, CSS styling, and JavaScript logic — no external dependencies, no server, no database. You download one file and it runs anywhere, with no build step or hosting account required.

These quizzes support multiple-choice questions, true/false, short answer fields, progress bars, per-question feedback (green for correct, red for incorrect), final score calculation, and a results screen. Everything runs client-side. You can add a timer, shuffle question order, or restrict retakes — all by including those requirements in the prompt.

What AI does not produce well in a single pass: quiz formats that require server-side response collection (like Google Forms does), conditional branching based on prior answers, or persistent analytics across many respondents. For basic interactive quizzes that self-grade, the single-file HTML approach handles everything.

How do you prompt ChatGPT or Claude to generate a quiz?

A precise prompt produces a working quiz on the first try. A vague prompt produces something that mostly works and needs debugging. Include these five elements: topic, question count, format, interactive behavior, and output constraint. Each element removes a decision the AI would otherwise guess wrong, cutting the number of follow-up iterations to near zero.

Here is a prompt that works on both tools:

Create an interactive quiz about the water cycle for high school students.
10 multiple-choice questions, 4 answer options each.
Highlight the correct answer in green after each selection.
Show a progress bar at the top.
Display the final score and a short message at the end.
Output as a single self-contained HTML file — all CSS and JavaScript inline, no external libraries or CDN links.

The “single self-contained HTML file” instruction is the critical constraint. Without it, ChatGPT may output code that references external CSS files or CDN libraries, which break when you open the file in a browser without a web server. Self-contained means one file, no dependencies, works offline.

Claude renders the quiz as a live artifact in the chat panel so you can interact with it before downloading. ChatGPT returns a code block — copy and save it as quiz.html, then open it in your browser to test.

Can you share a Claude artifact quiz directly from the chat?

Yes, and it is the fastest zero-setup option. Click the Share button in the artifact panel, then click Publish. Anthropic’s documentation confirms that anyone with the link can view and interact with the published artifact without signing up for Claude. They are only prompted to create an account for advanced features like remixing the artifact.

This is the right choice when you want a quick link for immediate use and do not plan to revise the quiz.

There is one hard constraint to know before you rely on it: the same documentation states that once you unpublish an artifact, you cannot publish that same artifact again. If you take the quiz down to fix a question and try to put it back at the same address, you cannot. You would need to publish a new artifact, which produces a new URL — breaking every link you already shared.

For a quiz you plan to update, downloading the HTML and publishing it elsewhere gives you control over the URL lifecycle.

How do you get the HTML file out of Claude or ChatGPT?

From Claude: In the artifact panel, click the download icon (top-right corner of the panel). Claude saves the file directly to your downloads folder as an HTML file. You can also click the copy icon to copy the source and paste it into a text editor.

From ChatGPT: ChatGPT returns quiz code in a code block. Click the “Copy code” button in the top-right corner of the code block. Open any text editor (Notepad on Windows, TextEdit on Mac, VS Code on any platform), paste the code, and save the file with a .html extension — for example, water-cycle-quiz.html.

Open the saved file in your browser before publishing. Click through the entire quiz: verify that answer selection works, score increments correctly, feedback appears, and the results screen shows at the end. If something is broken, paste the error or describe the bug in a follow-up message — “the score always shows 0 even when I answer correctly” is enough for the AI to identify and fix the problem.

Where can you publish an HTML quiz file and get a shareable link?

Several options exist. They differ on three things that matter for a quiz: whether the URL is permanent, whether you can update content in place without changing the link, and whether publishing requires a git repository or build pipeline. For a single HTML file, the last two are the deciding factors.

Method	URL permanent	Update in place	Requires git	Programmatic publish
CLI publish tool (e.g. dropthis)	Free: 7 days / Pro: permanent	Yes (update command)	No	Yes
Claude’s built-in Publish	Yes, but cannot republish after unpublish	No	No	No
Netlify drag-and-drop	Yes	Yes (re-upload same site)	No	Yes (via CLI/API)
surge.sh	Yes	Yes (redeploy same domain)	No	Yes (npm CLI)
GitHub Pages	Yes	Yes (git push)	Yes	Yes (git)

For a single HTML file with no build step, a CLI publish tool is the simplest path. One command, one URL:

dropthis quiz.html
# → https://your-quiz.dropthis.app/

The link works immediately. On a free plan, the drop stays live for 7 days. On a paid plan, it is permanent. The command returns a drop ID — keep it. You will need it to update the quiz later without changing the URL.

Netlify drag-and-drop and surge.sh are also solid options for permanent free hosting, though they require more setup than a single CLI command. Netlify’s free tier includes 100 GB bandwidth and 10 GB storage. surge.sh offers unlimited sites with a custom subdomain on a free plan. Neither has a native “update in place from a file” command as direct as the dropthis workflow.

How do you update the quiz without creating a new URL?

This is the question no other tutorial answers — and it matters most in the real use case: you publish a quiz for a class, students start sharing the link with each other, then you find a typo in question 4. You need to fix it without invalidating every link already in circulation.

The answer depends on how you published the quiz.

If you used a host that supports in-place updates, go back to the AI chat, make the correction, download the new HTML, and re-upload it under the same identifier. With the dropthis CLI:

dropthis update-content <drop_id> quiz-v2.html

The URL does not change. Anyone who loads the link after the update sees the corrected quiz. Anyone who was mid-quiz when you updated finishes on the old version (cached in their browser tab) — a non-issue for most quiz scenarios.

If you used Claude’s built-in Publish, you cannot update in place. Editing the artifact in chat creates a new version; it does not update the published URL. Your options are: share the new artifact link (breaking existing links), or switch to a hosting workflow that supports updates.

If you used GitHub Pages or Netlify, push the updated file to your repository or re-deploy the site. The URL stays the same; deployment takes a minute or two to propagate.

The update-in-place workflow is the reason to download the HTML rather than rely on Claude’s Publish button for anything you expect to revise. A URL you control and can update is more durable than a URL tied to a specific artifact version in a chat session.

Which path is right for your use case?

The right choice comes down to three questions: do you expect to update the quiz, do you need it to last beyond a few days, and do you want it independent of any AI platform’s infrastructure? For most practical scenarios — a class quiz, a team trivia night, a client demo — the answer to all three is yes.

Use case	Recommended path
Quick demo or one-time share	Claude’s Publish button
Classroom quiz, may need corrections	Download HTML, publish with update support
Permanent quiz for a course or website	Download HTML, publish to permanent host
Embed quiz in your own site	Paste the HTML into your site’s CMS or page editor
Collect responses with analytics	Use a quiz platform (Typeform, Tally, etc.)

For most people reading this, the classroom-quiz scenario is the right default: publish the HTML to a URL you control, keep the drop ID or site identifier, and update in place when the questions change.

The quiz-to-URL workflow mirrors the broader pattern of AI-generated artifacts: the AI does the hard work of building something functional; the missing step is always getting it to a URL that lasts and behaves predictably. The approach here — single HTML file, static host, update command — applies the same way to AI-generated games, landing pages, and any other self-contained output a chat tool produces.

AI quiz generation removes the barrier of writing questions and building a scoring engine. Publishing removes the barrier of “I have this file but nowhere to put it.” The combination means a working, shareable quiz from a single prompt in under ten minutes — no quiz platform account, no credit card, no build pipeline.