How to Share a ChatGPT Canvas as a Link (Behind Company Auth)

Why a Canvas is awkward to share

Canvas is the side panel where ChatGPT writes a document or app you can edit in place. When it generates an interactive piece (a dashboard, a chart, a small HTML app), you have something genuinely worth sending to other people. Then you hit the wall.

There's no "get a link to this Canvas" button. ChatGPT's share feature shares the conversation, not the artifact. So the recipient gets your prompts, your dead-ends, and the back-and-forth, with the Canvas rendered inside ChatGPT's own chrome. That link is also public by default and aimed at anyone, which is the wrong shape for internal work.

What you want is the opposite: just the finished artifact, at its own URL, visible to the right people, rendered exactly as built.

What you actually want to send

For a Canvas you're sending to colleagues, the bar is:

• Only the artifact. Not the conversation, not the prompts.
• Only the right people. Verified company identities, not the open internet.
• Interactivity intact. Charts, filters, and scripts still work, not a flattened screenshot.
• A permanent URL. One link that keeps working and that you can update in place.

Here are the realistic ways to get there, worst to best.

Get the Canvas out of ChatGPT first

Whatever you do next, start by pulling the artifact out. In Canvas, open the code/source view and copy the full HTML (or use the copy action on a code Canvas). That gives you a standalone file you control, independent of the conversation. Everything below works on that file.

Option 1 - ChatGPT's share link

Fast, but it shares the conversation, not the Canvas. Recipients see the whole thread, the link is public, and interaction still happens inside ChatGPT. Fine for showing a teammate how you got somewhere. Wrong for sending a clean deliverable to people who shouldn't see your prompts or shouldn't be on a public URL.

Option 2 - Paste into a doc or email

Copy the rendered text into Google Docs, Notion, or an email. Works for plain prose. It destroys anything interactive: a Chart.js dashboard or a script-driven layout becomes static text or nothing at all. Use it only when the Canvas is purely words.

Option 3 - Deploy it yourself

Push the HTML to Vercel, Netlify, or GitHub Pages. You get a real URL with the interactivity preserved. The cost is setup: a git repo, a project, a build step, and then a separate paid add-on if you want it restricted to your company rather than public. For one artifact you want a handful of colleagues to see, it's a lot of machinery, and the default result is still a public page.

Option 4 - Publish behind company auth

Take the Canvas HTML and publish it as a gated artifact. This is the one that matches all four requirements at once:

dsp publish ./canvas.html

You get back a permanent URL. A teammate clicks it, signs in once with their company Google or Microsoft account, and sees just the artifact with every chart and control working. No conversation, no public exposure, no accounts to create. Because pricing is flat with unlimited viewers, you can send it to the whole team without a per-seat bill.

Two things this adds that a plain deploy doesn't. First, inline comments: reviewers leave feedback pinned to the artifact, and you (or the model) read it, fetch the current version, and ship an update with the artifact ID and base version. That closes the loop between the people reviewing and the agent that wrote it. Second, it works for output from any tool, so the same workflow covers a ChatGPT Canvas, a Claude artifact, or a v0 prototype.

Not in a terminal? Publishing also works straight from an AI assistant over MCP. Say "publish this behind company auth" and the display.dev tool returns the gated URL in the conversation.

Picking the right option

Quick rule: a teammate who's allowed to see your prompts, native share link. A clean internal deliverable, publish it behind company auth. A static text document, a doc is fine.