Remove pages from GenSearch results
You can tag a single page in the article's Page Settings, or use the Page Classification Manager to apply the tag to multiple pages.
Use cases
Because semantically related content is returned from all pages for GenResponses, consider adding the llm-no-index tag to remove certain content from your GenSearch results. This helps prevents the LLM from generating incorrect or conflicting responses.
GenSearch respects user and page permissions. The focus of this article is on content usefulness, not privacy and permissions.
- Time- or status- based documentation
- Historical, legacy, or archived content
- Pages about deprecated or unsupported services or solutions
- Multiple variants of the same policy or procedure
- Conflicting instructions across different departments
- A/B testing documentation
- Regional variations that contradict each other without proper context
- Content around upcoming features, or features that are not available to all users yet
- Comment threads or forum discussions, which may be outdated or conflict with authoritative documentation.
Examples
Release Notes
Release Notes document recent feature updates, but an LLM does not understand that newer information should be prioritized as truth over older information.
Example: A 2019 release article says that you are using Internet Explorer, but in 2023 you moved to Microsoft Edge. A user asks GenSearch, "What browser does Company Z use?" and the response is, "Company Z uses Internet Explorer" or "Company Z uses Internet Explorer and Microsoft Edge."
Archived content
GenSearch cannot attribute the same lack of importance to archived content that a human would. If archived content is relevant to a search question, it may be returned in a GenResponse.
Example: You have archived pages about a deprecated feature. A user asks a question about how to use the legacy feature, and GenSearch walks them through the steps; the user cannot follow the steps in their more modern UI, and they become frustrated with your product and with generative AI. They do not know why GenSearch would return "useless" archived content.
Versioned content
Content about different versions of a product or service will not be recognized by the LLM as a separate version.
Example: Product A requires a new password every 30 days. Product B does not require password updates. A user asks GenSearch, "How often do I have to update my password?" and the response is, "You must update your password every 30 days."