Search testing

Before the LLM goes live, you can preview using the search testing feature to confirm the relevance and content of the data the LLM retrieves.

Why do search tests matter?

  1. Ensure response accuracy

  • The quality of the AI assistant's responses directly affects customer satisfaction

  • Incorrect or irrelevant answers can damage brand trust

  1. Prevent information omission

  • Prevent important product details or safety instructions from being overlooked

  • Optimize the efficiency of knowledge usage

  1. Improve answer consistency

  • Ensure the same question receives a consistent answer at different times

  • Avoid response instability caused by knowledgebase updates

With search tests and snippet preview, you can instantly view the quality of the data snippets the AI assistant uses to ensure your Q&A standards.

How to get started?

  1. Go to the knowledge page, select the knowledge base you want to view and enter settings

  1. Once inside, click Search Test to enter the page

Type what you want to search in the chat box: “What should I pay attention to when choosing a campsite?” After pressing search, the AI assistant will search the most relevant documents based on your query.

  1. View search results

On the right, the snippets the AI assistant will retrieve when answering will appear. As shown, there are 12 snippets in total (for retrieval snippet count settings, please refer to:How to create a knowledge base: Basic setup)

On the right are the retrieval contents used to answer this question. Each snippet will show:

  • Word count: number of words covered by the snippet

  • Hit count: the number of times the AI assistant referenced this snippet when answering (only counts formal Q&A; test Q&A are not included in hit count calculations)

    A higher hit count means users ask more questions related to that snippet.

Search tests record past test Q&A. You can quickly click history to retest, ensuring the same topics maintain or improve reference quality.

Check snippet quality

Through search tests, you can determine which snippets the AI assistant uses during reasoning and then check whether those snippets need content optimization, such as:

If you find many blank snippets in the search results and those blank snippets all belong to "Beginner Camping.pdf" this file, then you can check whether the contents of "Beginner Camping.pdf" are all correct, andupdate them in real timebefore going live.

Search tests help you

Regularly test important questions

  • It is recommended to frequently test core business-related questions

  • Retest after knowledgebase updates

  • Record and compare improvements in test results

Identify content gaps

  • Discover knowledge gaps: test common questions to find queries that return no hits at all

  • Example: Testing “tent waterproof rating description” finds no related snippets → need to add waterproof specification documents

Assess content completeness

  • Check answer depth: confirm whether the AI's answer covers all aspects of the question

  • Example: Query “camping safety precautions” → check whether it covers weather, wildlife, fire safety, and other aspects

Question testing tips

You can:

  • Ask the same question in different ways

  • Test complex scenarios and multi-condition queries

  • Verify retrieval effectiveness for technical terms and product names

Using these testing techniques, comprehensively evaluate the accuracy and completeness of responses.

Last updated

Was this helpful?