Generative AI pulls information from your help center and connected knowledge sources to provide answers to your users. Content is chunked and stored in a database for generative AI, so it can find the best matching content to generate replies.
Generative AI works best when content is clear, concise, and complete. If you’re using generative AI features, you can optimize your articles to improve the quality and accuracy of generative responses.
This article contains the following topics:
Understanding how knowledge sources are used by generative AI
For generative AI to use a knowledge source, the text is broken into chunks. Chunking is the first step in retrieval-augmented Generation (RAG). RAG lets a large language model (LLM) use information from connected knowledge sources.
These chunks are stored in a database that’s organized by semantic meaning. When a user message is sent, the meaning of that message is compared with the meaning of the stored chunks. Generative AI then uses the best matching chunks to generate a response.
- Knowledge content is imported and split into chunks.
- Each chunk is stored with a numerical vector, representing the semantic meaning of the chunk.
- Generative AI compares the user message with the stored vectors.
- Generative AI uses the matching information to generate the reply.
Text chunks created from your knowledge content are the source of generative AI, so if the content is clear and complete, the results will be better.
Best practices for optimizing content for generative AI
Generative AI works best when it draws from content that is clear, concise, and complete. Good preparation helps generative AI provide faster and more accurate responses.
Most of these guidelines also apply to good writing, in general, but you might find you want to adjust some of the guidelines to meet the needs of your human audience.
Best practices for writing content for AI use
Use these best practices for writing content:
- Ensure content is focused on the topic. Completely cover a topic and avoid covering multiple, unrelated topics.
- Provide complete answers. Each article should directly answer a user’s question, not imply an answer, and provide any context the user needs.
- Create self-contained content. Each article should contain all the information the user needs and avoid linking out.
- Keep wording clear and concise. Paragraphs should be short and focused, with sentences that are direct and to the point.
- Eliminate redundancies. Remove any duplicate articles or conflicting information.
- Define terminology. Explain terms in full and spell out acronyms on first use.
- Use wording your users are familiar with. Use wording that your users’ use, and that is used in your product or service.
- Avoid vague pronouns, such as “it” or “they.” Repeat the noun instead of using a pronoun that relies on earlier context.
Best practices for structuring and formatting content for AI use
- Structure content with headings. Use clear headings and subheadings to break content into logical sections.
- Use structured lists. Use bullet points for facts or tips and use numbered lists for procedures.
- Avoid nested instructions. If multiple solutions exist, present each as a separate instruction rather than sub-steps within a broader step.
- Include complete instructions. List all the steps needed to answer the question or solve the issue.
- Repeat the question or topic from the title. This ensures that context isn’t lost during chunking, and avoids separating questions from their answers.
- Add written text for pictures. Include text explaining any images, videos, or diagrams. Generative AI interprets text only.
- Avoid using tables. Comparison tables with minimal text can work, but in general, it’s harder for an LLM to understand information in tables than in sentences.