---
id: "en/ai-search"
title: "AI search"
language: "en"
description: "AI search is used to retrieve corporate documents, materials and knowledge within permission scope using natural language."
canonical_url: "https://www.babel.cc/public/babelwiki/en/ai-search.html"
markdown_url: "https://www.babel.cc/public/babelwiki/ai/en/ai-search.md"
alternates: {"en": "https://www.babel.cc/public/babelwiki/ai/en/ai-search.md", "zh": "https://www.babel.cc/public/babelwiki/ai/zh/ai-search.md"}
reading_note: "文档记录不等于当前部署能力；以实际版本、授权模块与配置为准。保留原文的可选、限制及待确认说明，勿补猜接口。 / Documentation is not a guarantee for every deployment. Preserve optional-module requirements, limitations and unresolved API contracts."
---

# AISearch

AI search is used to retrieve corporate documents, materials and knowledge within permission scope using natural language.

![AI Search](https://www.babel.cc/public/babelwiki/assets/images/ai-search.webp)
*Interface diagram of AI search.*

## First task: verify an answer against a policy document

**Prepare:** Choose a policy or guide you can open directly. Uploaded BabelBird files that AI can read are automatically ingested and vectorized for AI assistance, AI search and image search. After upload, test a question with a clear answer in the original. An empty result while ingestion is still processing does not mean the file has been lost.

1. Open **Zhichao AI** in the left sidebar and select **Intelligent AI Search (智能AI搜索)**. The screenshot shows this entry, the question field, and **Reference Sources (参考来源)** below the answer.
2. Ask a specific question about your document, such as “Which supporting documents does this expense policy require? Include the sources.” Wait for the answer to finish.
3. Check Reference Sources and open the cited file or its folder. Compare the file version and original text with the answer. Ask a follow-up question about any missing conditions.
4. **Completion check:** You can locate the source, and the answer's requirements, conditions and exceptions match it. Fluent wording alone is not verification.

**If no useful answer appears:** Open the original file to check access, then ask the administrator to check the retrieval scope, indexing completion and AI service status. If citations are missing or do not support the answer, locate the original with file-name or full-text search and verify it directly.

## Function description

AI search is oriented to enterprise information question and answer and knowledge retrieval. Users can input natural language questions, and the system will retrieve relevant documents within the permission scope and combine them with large models to generate answers. It is suitable for answering questions such as "What does a particular policy require?" "Where are the design materials for a certain project?" "What are the key terms in a certain customer contract?" "What are similar contents in certain types of documents?"

## Main abilities

- Study documents, knowledge materials and file contents within the designated scope of the enterprise.
- Determine the retrieval range based on user permissions to avoid unauthorized responses.
- Supports follow-up questions, users can continue to delve deeper into the last answer.
- The answer gives the source of the information, making it easier for the user to return to the corresponding file for verification.
- Supports recognition of text in pictures and can be combined with OCR and multimodal capabilities.
- Search results can include AI-generated content, or go back to the original file to verify the source.
- Complementary with full-text search, tag, description, annotation, file name search and other capabilities.

## Applicable scenarios

- Questions and answers about corporate policies, processes, contracts, project materials, product descriptions and customer service materials.
- Data retrieval for R&D, engineering, design, legal, administrative, human resources and other departments.
- Quick location of multi-file, multi-project, and multi-version data.
- Allow new employees to ask questions to understand the company's information, rather than rummaging through directory after directory.


## Permissions and usage boundaries

For shared requirements on source authorization, model services and human review, see [Zhichao AI Overview](https://www.babel.cc/public/babelwiki/en/ai-overview.html). The completion checks for this task do not replace your organization’s access approvals.

## Related information

- [Permission system](https://www.babel.cc/public/babelwiki/en/permissions.html)
- [Security and Audit](https://www.babel.cc/public/babelwiki/en/security.html)
- [FAQ: Deployment, AI and technical issues](https://www.babel.cc/public/babelwiki/en/faq-technical.html)
- Public introduction: <https://zhuanlan.zhihu.com/p/2052372550513072059>