The audit that happens as part of AI Readiness classifies issues using a set of content issue types.
Issue type severity
The definitions for each issue type are listed in this section in severity processing order.
- Conflicting Facts (Contradiction)
-
The system has identified multiple instances of content that answer the same question (e.g., differing dates, policies, or code steps), this may occur within a single page, or involve multiple pages.
- The suggested revision should indicate what information is in contradiction, and the editor can identify the correct information and update or delete the incorrect information.
- Incomplete Information (Content Gap)
-
The system has identified content on multiple pages that provides answers to a question, but the different pages do not provide the same information. For example, one page may be missing some information for the answer that the other pages provide.
- The suggested revision should indicate what information is missing, and the editor can update the content on those pages with the missing information.
- Ambiguous Content (Answer Quality)
-
The system has identified content on multiple pages that provides answers to a question, but the different pages do not provide the same level of information. For example, one page may provide detailed information including restrictions or limits, while another page provides a generic statement without mentioning them. If a customer just used the information from the second page, the customer might get misleading information without the full context.
- The suggested revision should indicate what information is missing and how to incorporate it into the page so that both pages provide the same level of information.
Issue detection behavior
Once the audit flags an issue type, it stops checking for other issue types ranked lower in severity.
For example, if the audit finds a Contradiction issue type, it will not continue checking for Content Gap or Answer Quality issue types.
Why does the product stop checking after the first issue found?
Checks are performed in order of severity: conflicting content is identified first, followed by incomplete content, then ambiguous content. When an issue is found, further checks are skipped.
This is intentional. There is no value in checking whether content is complete or unambiguous if it is already contradicted — contradicted content may be incorrect, so completing or clarifying it would mean building on a potentially false foundation. Similarly, if content is incomplete, adding nuance or resolving ambiguity in what exists does not address the more fundamental gap.