Google Search

Google Tells Platforms to Keep Link-Shared Pages Out of Search

View as Markdown RSS
The title and opening paragraph of Google's guide, Prevent user-generated spam on your site and platform. The words 'and tools that let users publish shared pages, generated content, or web apps' after 'Hosting platforms', and the words 'pages or' before 'sites', are highlighted in yellow.
The introduction to Google's guide to preventing user-generated spam. The words added on 26 October 2026 are highlighted.

On 26 October 2026 Google updated its Search Central guide to preventing user-generated spam, which sets out how any site or platform that accepts content from its users can keep spammers out, from abuse policies and moderation to blocking automated sign-ups. Its introduction named comment forms and hosting platforms as the openings spammers exploit, and it now also names tools that let users publish shared pages, generated content or web apps. A new section says that pages made for sharing, collaboration or utility, rather than to be found in search, should be kept out of the index by default.

Part of Google's guide to preventing user-generated spam. Between an existing paragraph on nofollow and ugc links and the existing Use manual approval heading, the new Only index content meant for Search section is highlighted in yellow: services should consider whether shareable pages, generated outputs or interactive tools are meant to be found through search, and pages made for direct sharing should get a noindex rule or a robots.txt block by default.
Google's guide to preventing user-generated spam. The "Only index content meant for Search" section added on 26 October 2026 is highlighted.

Google also added two monitoring points: verify the site in Search Console so manual action and security emails reach you, and watch the Page Indexing and Performance reports for spikes in indexed pages, or in queries unrelated to what the platform does. Its changelog says the aim is to keep user-created pages out of the index and catch spam earlier.

What kind of pages is Google describing?

Google has not said what prompted the change, but AI chat products already make pages of this shape: a shared conversation or a published artifact sits at a URL on the AI company’s own domain, open to anyone with the link. Such pages have reached Google’s results by accident twice. In 2025, ChatGPT users could tick “Make this chat discoverable” on a shared link, until OpenAI removed the option on 1 August after indexed chats were found containing personal information. In July 2026, shared Claude conversations surfaced in Google the same way.

Both were privacy incidents, but the same feature is an opening for spam, and the wording of the guide’s introduction shows the change. It used to say hosting platforms were open to abuse because spammers may create a large number of sites. It now adds tools that publish shared pages, generated content or web apps, and says spammers may create a large number of pages or sites. Spam of that kind needs no site of its own: a page on someone else’s domain puts the spammer’s content and links on ranking signals they did not earn.

Google’s site reputation policy, its rule against third-party content published to ride a host site’s established ranking signals, lists sites designed for user-generated content among the cases it does not cover, so this is handled as user-generated spam, where the responsibility sits with the platform. That is why the new monitoring advice is aimed at platforms: it is the platform’s site that Google says may receive the manual action.

Should you noindex shared pages or block them in robots.txt?

Use noindex, and leave the pages crawlable. The new section offers a robots.txt block as an alternative, but Google’s robots.txt documentation warns that a blocked URL can still be indexed if other pages link to it, because Google never fetches the page and so never sees a noindex on it. That was the Claude problem in July: the share pages carried a noindex header on a path that robots.txt disallowed. claude.ai had removed the block by 30 July, and as of 26 October 2026 its share and artifact pages return an X-Robots-Tag: none header, which Google reads as noindex, nofollow.

If your product has share links, AI-generated outputs or user-built apps, serve noindex on those pages by default and set an alert on the indexed count in the Page Indexing report. Verify the domain, and any subdomain users publish on, in Search Console, so a manual action reaches you as an email rather than as a traffic drop. Anything that has to stay private needs a login, because neither directive keeps a page private.

Sources

More news