MakeItRun blog

Your Claude Artifacts Are Showing Up in Google

2026-07-28 · MakeItRun team

TL;DR. Yes — a published Claude artifact is a public URL the moment you hit Publish, no viewer account required, and Anthropic's own help docs say so plainly. This week a Reddit thread showed that some of those links, along with shared Claude conversations, are turning up in ordinary Google searches: business documents, spreadsheets, internal reports, things people published thinking "just a link for one person" rather than "public web page." Anthropic has since added the missing noindex signal and Google pulled the results it had — but the fix is forward-looking, not retroactive, and Bing's index lagged behind. Check your own in two minutes: search site:claude.ai/public/artifacts plus your name or business, then open your Claude account's list of published artifacts and unpublish anything you wouldn't want a stranger to find. One warning before you do: unpublishing is permanent. Read the whole thing before you click it, especially if you're not sure which artifact is which.

The Reddit thread that surfaced this — first flagged over the weekend and picked up within days by TechCrunch, VentureBeat, Axios, and Cybernews — wasn't a hack or a breach. Nobody broke into anything. People used Claude's Publish button and Share feature exactly as designed, and a chunk of what they published turned out to be a Google search away. That gap between what "publish" sounds like and what it actually does is the whole story.

What Publish actually does

Anthropic's own help center article is direct about it: publishing an artifact "makes your artifact publicly available. Anyone with the link can view and interact with it," and a visitor can do that "without signing up" — no Claude account needed on either end. That's the entire mechanism: one click turns a chat-window artifact into a standing claude.ai/public/artifacts/… page that stays up until you unpublish it.

None of that is a secret, and none of it is a bug. It's also not the same thing most people picture when they hear "share a link with someone." A link that anyone can open, with no login and no expiry, is a public web page. Whether a search engine also happens to find it is a separate question — and this week's news answered it.

What happened this week

According to the reporting, the Reddit post showed that plain site: searches on Google surfaced hundreds of shared Claude conversations and published artifacts — enough that the thread reportedly gathered several thousand upvotes within hours. The categories described across multiple outlets: financial models and payroll spreadsheets with real names attached, CRM exports, unreleased product plans, internal company documents, medical and legal material, and administrative records like a university's internal academic calendar. We're not linking to any of it, and we're not naming anyone in it — the pattern is the story, not any individual page.

The mechanism, per Cybernews's reporting, is a genuinely instructive detail: Claude's shared and published pages were blocked from crawling in robots.txt, but they were missing the separate noindex signal. Those aren't the same protection. A robots.txt disallow tells a crawler not to fetch a page's contents — it doesn't stop a search engine from listing the bare URL if that URL shows up somewhere else on the open web, which is exactly how a link posted to a forum or social feed gets picked up. Only an explicit noindex, seen on a page a crawler is actually allowed to visit, reliably keeps a URL out of results. Anthropic added that signal and Google had removed the flagged results by the next day; Bing's index reportedly lagged, and anything already scraped or archived elsewhere isn't something a noindex tag can undo.

Why this catches non-technical users specifically

Nobody who builds a client proposal, an invoice calculator, or an internal price list in Claude is thinking about search engine indexing. They're thinking "I need to send Dana this link." Publish does that job perfectly — right up until the link outlives the one conversation it was made for: it gets pasted into a group chat, forwarded into a doc, dropped into a forum thread asking for feedback, and now it's sitting somewhere a crawler can see it. None of those steps require anyone to do anything careless. That's the actual risk, and it's also why "just unpublish it if that happens" isn't a full answer — Anthropic's help center states plainly that once you unpublish an artifact, you can't publish that same one again. You'd rebuild it from a new copy, at a new link, and redistribute that instead. There's no undo button that gets the old link back under control; there's only forward.

How to check what you've actually published

Search for yourself. Try site:claude.ai/public/artifacts combined with your name, your business name, or a phrase you know appears in something you published. Repeat it on Bing — the two indexes don't stay in sync, and this story is part of why that matters right now.
Open your own list. Claude keeps a record of what you've published and shared under your account settings. Most people find more live links than they expected — old drafts, test versions, things a coworker asked for once and never followed up on.
Unpublish anything sensitive, carefully. Client data, internal pricing, anything with someone else's name in it — take it down. Just know going in that the action is final for that artifact; if you'll need the content again, save a copy of the source before you unpublish.

The honest fix, and where we land

This isn't a "switch to us and the problem disappears" post — the same failure mode is possible on any platform that doesn't treat search indexing as a decision the publisher makes on purpose. What we can say precisely, because it's in our own code: every page published through MakeItRun ships with an x-robots-tag: noindex, nofollow response header by default. A page only becomes indexable when its owner explicitly turns on a "Make this page findable on Google" toggle in the dashboard — and that toggle stays off unless a page has cleared moderation, so a flagged page can't be made indexable even by choice. Anonymous, no-login publishes can never be flagged indexable at all. Beyond indexing, each page also picks an access mode — a public link, signed-in-only, an allowlist of specific emails or domains, or fully private — so "who can open this" and "can a search engine find this" are two separate settings, not one checkbox standing in for both.

The free tier caps at three published pages and shows a small attribution bar — that's what pays for the bandwidth a free page uses; paid tiers remove it. A QR code for any published page is a real download from the dashboard, not something stamped onto the page automatically. None of this makes the underlying judgment call — what should and shouldn't be public — for you. It just means the default doesn't decide it for you either.

FAQ

Are Claude artifacts public by default?

An artifact only becomes public when you click Publish. Before that, it lives inside your private conversation. Once published, it's a standing public URL — no account required to view it — until you unpublish it.

Can Google actually index a published Claude artifact?

It can, if the link ends up somewhere a crawler can see it — a forum post, a social share, a link-in-bio page. Anthropic's own robots.txt reportedly blocked crawling but was missing a noindex signal, which is why some published pages and shared conversations showed up in Google results this week. Anthropic has since added noindex; Google removed the flagged results the following day.

If I unpublish my artifact, does it come out of Google?

Unpublishing takes the live page down, which should eventually clear it from search results as Google recrawls. It doesn't retroactively remove copies that were already scraped, archived, or indexed elsewhere in the meantime, and Anthropic's help center confirms you can't republish that same artifact — you'd need to rebuild it as a new one.

Does MakeItRun index my published pages on Google?

Not unless you turn it on. Every page defaults to a noindex response header; it only becomes indexable if you explicitly flip the "Make this page findable on Google" toggle for that page, and only if it's cleared moderation.

Published something you meant for one person?
Get a link where "public" and "findable on Google" are two separate choices.
Get your link →