Tianmu Intelligent Recognition System

The "Tianmu" Intelligent Recognition System is an AI-generated content detection tool developed by the State Key Laboratory of Communication Content Cognition of People's...

Screenshots

A well-written copy, a neatly structured academic paper, or a polished industry report may be the “creation” of a large language model in just a few seconds. As AI-generated content floods the Internet, determining whether text is human-made or machine-generated has become an unavoidable new challenge for industries such as academia, publishing, and education. The Tianmu Intelligent Recognition System is an AI content detection tool developed to address this pain point. It is deeply customized and researched by the State Key Laboratory of Communication Content Cognition of People’s Daily Online, advocating a new content risk control idea of “governing AI with AI”. It can accurately identify Chinese text generated by AI, detect deepfake content and trace its synthesis methods, ensuring the credibility of text and the authenticity of information.

Where is the official entry? First, get to know its “background”

Official website: https://tianmu.people.cn/

The developer of the Tianmu Intelligent Recognition System, the State Key Laboratory of Communication Content Cognition of People’s Daily Online, has long focused on enhancing content risk control capabilities through technological innovation. Leveraging the technical expertise of the national laboratory in the field of communication content cognition, the Tianmu system does not stop at “general detection” but is deeply adapted to the generation characteristics of mainstream AI large language models at home and abroad, so its performance in recognizing Chinese text is particularly outstanding.

For ordinary users, the experience threshold is not high: open the official website, register and log in to an account, then complete the entire process from uploading files to obtaining reports in the browser, without installing any client software.

Core Capabilities Breakdown: Accuracy, Capacity and Security Are All Indispensable

Whether a detection tool is reliable usually depends on three things: accuracy of judgment, ability to handle large documents, and data security guarantee. The Tianmu system has clear technical indicators supporting these three dimensions:

• High-precision Recognition: For Chinese text generated by major AI large language models at home and abroad, the recognition accuracy is as high as 93%, giving it a significant advantage in the field of content risk control;
• Large-capacity Detection: Supports up to 100,000 Chinese characters per text detection, so long academic papers, industry reports, and serialized manuscripts can all be processed at once;
• Data Privacy Protection: Does not retain any detection data, nor uses detection data for other purposes, so users do not need to worry about the risk of manuscript leakage;
• Highlight AIGC Paragraphs: The detection results will highlight the specific sentences and paragraphs generated by AI, making it clear where the problem lies;
• One-click Report Generation: Supports one-click generation of detection reports, which is convenient for archiving, keeping evidence, and sharing.

Why the 93% Accuracy Rate Is Worth Attention

The difficulty of AI detection lies in balancing “false positives” and “missed detections”: if the judgment is too strict, manually original content may be mislabeled as AI-generated; if the judgment is too loose, machine-generated text will slip through. The Tianmu system has achieved a 93% accuracy rate in Chinese text recognition, which means that in scenarios with strict requirements for conclusions, such as academic paper review and manuscript auditing, the detection results have high reference value and can significantly reduce the workload of manual review.

Govern AI with AI, which essentially means using tools that understand the generation principles better to identify the outputs of another batch of generation tools.

Practical Test Process: Get a Detection Report in Five Steps

The operation path of the Tianmu system is quite linear, so even first-time users can complete the whole process smoothly:

1. Registration and Login: Visit the official website, register an account and complete login;
2. Upload Files: Upload the text files to be detected; currently, two common formats, DOC and TXT, are supported;
3. Automatic Detection: The system automatically performs text analysis, no need to manually set parameters;
4. View Results: View the overall conclusion on the results page and locate the AI-generated paragraphs that are highlighted;
5. Download Report: Generate a PDF-format detection report as needed, and download or share it with others at any time.

In the whole process, the upload format is one of the few details that need attention: if your manuscript is in other formats, it is recommended to save it as DOC or TXT before uploading, which can avoid unnecessary format conversion problems.

Who Uses It and How to Use It: Inventory of Three Typical Scenarios

Academic and Educational Scenarios

Screening AI-generated content from students’ submitted papers and course assignments is the scenario with the most concentrated current demand. Teachers and educational institutions can use the highlight positioning function to quickly determine which paragraphs are suspected to be machine-generated, then confirm with manual interviews, making academic integrity review based on solid evidence.

Content Creation and Media Scenarios

Editors can use the Tianmu system to verify the originality of submitted manuscripts during review, and platform operators can also incorporate it into the content risk control process to filter low-quality, purely machine-assembled submissions, maintaining information authenticity from the source.

Enterprise Application Scenarios

The Tianmu system provides enterprise-level services and opens API interfaces, making it suitable for embedding AI detection capabilities into their own business systems, such as manuscript management platforms, online education backends, etc., to achieve batch and automated content review. It should be noted that the system currently supports Chinese text detection, and may expand to other languages in the future.

Quick Overview of Common Questions

• Will detection data be saved? No. The system does not retain or reuse any detection data;
• What about long documents? Supports up to 100,000 Chinese characters per time, so most papers and reports do not need to be split;
• Can it connect to my own system? Yes, enterprise users can integrate the detection capability through API interfaces.

Several Practical Tips: Make the Most of Tianmu

• When it comes to formal purposes (such as academic review, manuscript final review), it is recommended to generate and save a PDF detection report to form a traceable written basis;
• After getting the detection results, do not only look at the overall conclusion, but should check each paragraph in combination with the highlighted paragraphs; combining AI detection conclusions with manual judgment is more reliable;
• For users with high data confidentiality requirements, you can focus on Tianmu’s data strategy of “no retention, no reuse”, which is also the key point that distinguishes it from many similar tools;
• For teams with batch detection needs, it is recommended to consult enterprise-level services and API integration solutions first, which is far more efficient than uploading each manuscript manually.

From the 93% Chinese recognition accuracy rate, to the 100,000 Chinese characters per detection capacity, and then to the privacy commitment of no data retention, the Tianmu Intelligent Recognition System has turned AI content detection into a standardized and implementable capability. Whether you are a teacher who needs to verify papers, an editor who reviews manuscripts, or an enterprise that wants to build a content risk control system, you can upload a manuscript through the official website to test the recognition effect of this “govern AI with AI” system yourself.