AI Detection Assistant - Free High-Accuracy Chinese AI Rate Detection

The Specific Problem for Engineers

Within Chinese content creation and review pipelines, a persistent challenge is the need to quickly and reliably determine if a block of text is predominantly AI-generated. For engineers and developers handling academic paper excerpts, marketing copy, or social media posts, the absence of a lightweight, reliable, and free detection tool means relying on manual judgment or expensive commercial services, which is inefficient and costly in automated workflows. Existing solutions may lack robust support for Chinese corpora or be difficult to integrate into custom scripts.

How the Skill Works: Core Capabilities & Key Steps

The core capability of this skill is assessing the probability of AI generation for short texts, returning structured results via a simple API call. Its operation revolves around several key steps:

1. Text Preprocessing & Validation: The skill first validates the input text. It strictly enforces a length constraint between 300 effective characters (whitespace excluded) and 1000 characters (punctuation included). This ensures the analysis model receives sufficient, but not overly long, context for effective detection.
2. Calling the Detection Endpoint: The skill prioritizes using a wrapped Python script to initiate the request. The script handles common issues like CERTIFICATE_VERIFY_FAILED and can automatically fall back to the system curl if necessary. Upon a successful call, it parses the returned JSON data.
3. Result Parsing & Presentation: The response includes key fields such as avgAiProbability and avgHumanProbability (both probability values between 0 and 1) and a Chinese summary interpretation. The skill formats this data into clear percentages for the user, accompanied by a "for reference only" notice. For example, an AI probability of 75% indicates the model judges the text has a 75% chance of being AI-generated.
4. Error & Boundary Handling: The skill has explicit error handling logic for various failure scenarios, such as invalid word count, rate limiting (429), or server-side errors (5xx). This includes guiding users to correct the text, advising to wait and retry, or suggesting the full official web tool.

Applicable Boundaries & Important Notes

The tool's design boundaries are clear and must be observed:
- Strict Length Constraints: This is the core limitation. Input must be between 300 and 1000 characters. For text outside this range, the skill will truncate to the first 1000 characters for a preview detection and strongly direct the user to the complete website (https://www.ailv.run/). Do not attempt to detect long texts by repeatedly calling the short-interface.
- Rate Limiting: The calling frequency is approximately 10 times per minute. If a 429 error is encountered, read the retryAfterSec field, wait, and retry at most once more.
- Functional Positioning: It is a detection and assessment tool, not a rewriting or ghostwriting tool. The output probabilities and suggestions are for reference only; final text modification decisions should be made by the user.
- Privacy & Security: Text is used only for instantaneous detection and is not stored on the server. Additionally, the skill design prohibits storing or transmitting API keys anywhere, ensuring secure usage.