Published Claude artifacts are public. Here is how to make one private.

Are published Claude artifacts public?

Yes. Publishing toggles the artifact to public and produces a URL on claude.ai that anyone can open, with no login required. This is the intended behaviour: publish is the share-with-the-world button. What it is not is a private-sharing feature. There is no password, no expiry, and no view analytics in the native flow, and logged-in users can remix a published artifact into their own copy. This covers standard HTML artifacts; AI-powered or Live artifacts that call Claude at runtime still require the viewer to sign in.

One thing that does stay private: the surrounding conversation. Publishing exposes the artifact only, not your chat history or the prompts that produced it. That distinction matters less than it sounds, because the artifact usually carries the data it was built from.

Do Claude artifacts get indexed by Google?

They can. Anthropic's help documentation on publishing artifacts notes that published artifacts may be indexed by search engines. In practice that means a published artifact is not just reachable by people you sent the link to; it can be crawled, indexed, and surfaced in Google results to anyone searching. An unpublished artifact has no public URL, so it is not indexable.

This is the gap most people miss. "Share via link" feels private because you only sent the link to one person. But a public, indexable URL is discoverable independently of who you shared it with.

Can people without the link find my Claude artifact?

If it is published, potentially yes. Because the URL is public and may be indexed, someone who never received the link can still land on it through search. If the artifact is unpublished, it stays inside your account and is not reachable. There is no middle "unlisted but uncrawlable" state in the native publish flow, which is the gap a private host fills.

What is actually exposed

• The rendered artifact. Public to anyone with the link, and reachable via search if indexed.
• Whatever the artifact embeds. Agent-generated HTML routinely inlines API responses, customer rows, internal IDs, or prompt context. Publishing the artifact publishes that data too.
• A remixable copy. Logged-in users can fork a published artifact into their own account.

What is not exposed: your chat history and the conversation around the artifact.

How to make a Claude artifact private

There are two honest paths.

1. Do not publish it (or unpublish it). An unpublished artifact lives inside your Claude account and is not on a public URL. You can unpublish a version at any time, and per Anthropic's docs, once you unpublish a version you cannot republish that same version. The catch: pages already crawled can linger in search caches until they are re-crawled, so unpublishing is a cleanup, not a guarantee. And an unpublished artifact cannot be opened by someone without a Claude account.
2. Host the HTML privately. Copy the artifact HTML out of Claude (or have Claude Code write it to a file) and publish it to a private-by-default host. The result is a link anyone can open, no Claude account needed, that is not indexed and not discoverable by guessing.

The private-by-default fix

Stacktree inverts the default. Where Claude publish is public-and-maybe-indexed, Stacktree is private unless you choose otherwise:

• Unguessable URL. The token is the credential. No directory listing, no public index, nothing to find by guessing.
• Search and AI opt-out. Every site is served with an X-Robots-Tag opt-out, so even a leaked URL signals crawlers to skip it.
• Optional gates per link. Add a shared password, an @yourco.com email-domain gate, or an expiry, link by link.
• Replace-in-place. When the agent revises the artifact it can call update_site with the same slug, so the URL you sent stays valid while the content updates.
• No Claude account to view. Recipients open a normal link. They do not need a Claude account or a Stacktree account.

If your goal is simply to share an artifact with someone who lacks a Claude account, the step-by-step flow lives on the share Claude artifacts page. This page is about the privacy and indexing question specifically: what is public, what gets crawled, and how to keep an artifact off the public web.