Add an HTML attribute to prevent content from being indexed
There are times when sections of content within an article should be returned verbatim, and therefore should not be indexed by Kernels. This can include specific terms and conditions, regulatory requirements, or other content you do not want to be paraphrased or put into different verbiage.
Contact Expert to enable this feature on your site.
Use cases
- Prevent context confusion. Sensitive instructions, warnings, or disclaimers where paraphrasing could create liability or safety issues.
- Maintain legal accuracy. Contract clauses, terms of service, or compliance text that must appear exactly as written.
- Control brand consistency. Marketing copy and strategic content, product names, or slogans and taglines.
- Maintain code integrity. Code snippets, command-line examples, or configuration files where any modification breaks functionality.
- Ensure diagnostic precision. Error messages, log formats, or troubleshooting steps where exact wording is critical for user or agent recognition.
- Preserve historical accuracy. Quotes, citations, or versioned documentation where exact reproduction is required for reference purposes.
How to remove content from indexing
- Navigate to the HTML view of the page.
- Add the attribute
"data-llm-index"=falseto content that should not be indexed. - Save.
If "data-llm-index"=false is added to a page that has already been indexed, that piece of content is removed from the index.
Examples
The HTML for unindexed text will look like the following:
<p "data-llm-index"=false>Content that should not be indexed.</p>
Within a paragraph
Text as it appears on the page:
"To adhere to product guidelines, note the usage of certain terms and phrases. One important phrase we use is our tagline Where the right ones are left. This is always used in its entirety, and should never be abbreviated, paraphrased, or otherwise altered in our documentation."
HTML with the tagline removed from index:
Description of a table or image
Content as it appears on page:
"The following table is how the People Person's Paper People report our weekly paper sales, with an example:
| Date | Type | Reams sold | Cost per ream ($) | Total cost ($) |
|---|---|---|---|---|
| 8/20/2024 | White A5 24# | 3 | 2.45 | 7.35 |
This table should be attached to the weekly recap report you send to Kevin every Monday."
HTML with the table description removed from index:
