What is Zhuque AI? How to use the free AI content detector

Over the past year or so, the feeling of "I can't tell if this is real or AI-made" has become an everyday thing. A post that reads too polished to be true, an image that looks unreasonably perfect, a video of a public figure saying things they never said: all of it can be the work of AI. For content people, teachers, and editors, telling human-made from machine-made is no longer idle curiosity, it has become a genuine work requirement. This is exactly the gap Tencent's Zhuque AI wants to fill. Instead of asking you to guess with the naked eye, Zhuque AI returns a concrete probability score for whether a piece of text, an image, or a video was likely generated by AI. In this article, I'll dig into how Zhuque AI works, who it's for, how to use it well, and the limitations you should know before you trust its results completely.

What is Zhuque AI?

Zhuque AI (full name 朱雀 AI 检测, or Zhuque AI Detection Assistant) is an AI-content detector built by Zhuque Lab, the AI security lab under Tencent's security division, founded in 2019. Unlike tools that can only inspect one type of content, Zhuque AI is multimodal: it can check text, images, and video.

The way it works is easy to picture. The system is trained on a massive dataset of both human-made content (written, drawn, filmed) and AI-generated content, and from that it learns to spot "fingerprints" that are hard to catch by eye, such as unusual statistical patterns in wording, or the structural and texture errors that only AI tends to make in images. When you feed it a piece of content, Zhuque AI returns a percentage probability that the content is AI-made, along with the passages it flags as most suspicious.

Who Zhuque AI is best for

Teachers and lecturers

With the wave of students using ChatGPT for their assignments, teachers need a quick way to check the originality of what gets turned in. Zhuque AI lets you paste an essay directly or upload a file, then highlights the passages it suspects are AI, giving instructors a reference point for following up with students. Treat it as a grading aid, not as absolute proof.

Editors and journalists

In an era of fake images and deepfake video, journalists need to verify where an image came from before publishing. Zhuque AI's strength in image and video detection helps editors catch "too good to be true" images from Midjourney or Stable Diffusion early. That helps a newsroom avoid the risk of unintentionally spreading misinformation.

Content creators and community managers

If you run a fan page, a TikTok channel, or a community with a lot of user-submitted content, Zhuque AI helps you quickly filter out low-quality AI content or fake images before you approve them. That keeps your channel's quality up and lowers the risk of publishing something that breaks the rules. For a busy moderator, an automatic filter layer like this saves a fair amount of time.

AI development and training teams

For people working with data, accidentally mixing AI-generated content into a training set can "contaminate" a model and lower its quality. Zhuque AI helps screen that out and keep the genuinely human-made portion of the data. It's an underrated use case, but a valuable one for technical folks.

Standout features of Zhuque AI

Text detection in Chinese and English

Just paste a passage (or upload a TXT/DOCX file) and Zhuque AI scores the probability it was AI-written and highlights each suspicious section. It recognizes output from most popular models like GPT-4, Claude, DeepSeek, Gemini, and Hunyuan, across genres ranging from essays and news to fiction and official documents. Tencent puts text accuracy at over 92%.

High-accuracy AI image detection

This is Zhuque AI's strongest feature. The image detector alone was trained on more than 1.4 million samples and reportedly reaches over 95% accuracy, recognizing images from Midjourney, Stable Diffusion, DALL-E, Flux, and GAN-based models. It inspects tiny details like surface texture, illogical lighting, or abnormally repeating patterns that the human eye usually misses.

AI-generated video detection

Zhuque AI has extended its capability to video, helping detect clips produced by models like Sora, Hunyuan, and Hailuo. You can upload common formats like MP4 and MOV for the system to analyze. It's a notable step forward as deepfake video gets harder and harder to tell apart by eye.

Probability reports and visual flagging

Instead of giving a plain yes or no, Zhuque AI exports a report with a percentage AI-risk score for the whole piece of content. For text, the tool highlights each suspicious passage so you know exactly where to look again. This visual presentation makes the results readable even for non-technical users.

Benefits of using Zhuque AI

Free and ready to use

The biggest plus is that you don't pay anything for personal use, and there's no software to install. Just open a browser on your computer or phone, log in, and you're good to go. Compared with many overseas tools that charge by character count, that's a clear advantage.

Multi-format checks in one place

Instead of using one tool for text and another for images, Zhuque AI brings text, images, and video onto a single platform. That saves time and makes the verification workflow tidier. For content people juggling many types of material, it's a real convenience.

Results in a few seconds

Zhuque AI's processing speed is impressive, usually returning text results in a few seconds, with images only slightly slower. That means you can check a batch of content without much waiting. It's an important factor for anyone who needs to verify something in a hurry.

Respect for privacy

Per Tencent's policy, the content you submit for checking isn't stored permanently or used to train models. That gives you more peace of mind when checking sensitive documents. Even so, you should still save the report yourself as a screenshot for later reference.

How to use Zhuque AI, step by step

Step 1: Visit the website

Open your browser and go to https://matrix.tencent.com/ai-detect/. The interface is fairly simple: there's a paste and input box in the middle plus an upload area for images and video, so you can start right away without any complicated setup.

Step 2: Log in to your account

You can try it as a guest with a limited number of runs, but to get your full daily quota of checks, log in. Zhuque AI supports signing in with a phone number or an account in the Tencent ecosystem.

Step 3: Pick the content type to check

On the interface, choose the tab that matches what you want to inspect: text (文本), image (图片), or video (视频). Each type has its own input area.

Step 4: Add your content

For text, paste the passage directly or upload a TXT/DOCX file (up to about 2,000 Chinese characters per run). For images or video, upload the file from your device into the designated box.

Step 5: View and read the results

Hit the check button and wait a few seconds. Zhuque AI returns a percentage likelihood that the content is AI-made, along with highlighting of the suspicious passages (for text). It's worth screenshotting the report to keep on file.

Tips for getting the best results from Zhuque AI

Split long text to inspect it more carefully

Since each run only checks about 2,000 characters, break long documents into parts. That works around the character limit and helps you pinpoint exactly which section has a high AI score to revise. It also makes the results more reliable than scanning the whole piece in one rushed pass.

Prefer original, high-quality images when checking visuals

Zhuque AI detects AI images based on very small traces in an image's texture and frequency, so when you check, upload the original file rather than a screenshot or a heavily compressed copy. Re-capturing or compressing an image can wipe out those traces and make the result less accurate. The closer the image is to the original, the more reliable the result.

Treat the result as a reference, not a verdict

No AI detector is 100% accurate, and neither is Zhuque AI. So use the probability score as a signal to look more closely, rather than a firm conclusion that a piece of content is AI-made or human-written. Combining it with your own judgment gives the most accurate decision.

Zhuque AI pricing

Right now Zhuque AI follows a free-for-individuals, paid-for-business model. Note that Tencent hasn't published an official detailed price list on its website; the free-tier figures below are compiled from third-party guides and may change, so check the latest notice on the homepage.

• Free plan (individual):
• No cost, used directly in the browser, no installation
• Reference limit of roughly 20 text checks per day (some sources add roughly 20–30 image checks per day, but this figure isn't consistent across sources)
• Up to about 2,000 Chinese characters per text run
• Suited to students, content creators, and small teams that need daily self-checks
• Business plan (paid):
• Provides an API for integration, batch processing, and private deployment
• Suited to newsrooms, social platforms, or businesses that need large-scale content moderation
• Pricing isn't published publicly; businesses need to contact Tencent directly for a quote

Tencent hasn't opened a paid plan for individual users yet, so once you use up your free daily runs, you usually have to wait until the next day for the quota to reset.

Limitations of Zhuque AI

Limited language support beyond Chinese and English

This is the most important limitation for anyone working outside those two languages. Zhuque AI's text detection is optimized for Chinese and English, so if you paste in text written in another language, the result is close to unreliable. For content in other languages, look for a tool built specifically for that language rather than leaning on Zhuque AI.

Login and access friction outside China

As a China-domestic product, logging in is often smoothest with a phone number or an account in the Tencent ecosystem, which can be awkward for users abroad. Access speed from outside China can also be inconsistent at times. It's worth lining up a suitable login option in advance.

Limits on run count and length per check

The free version only allows about 20 checks per day, and each text run caps out at roughly 2,000 characters. If you handle large volumes of content or long documents, you'll have to split them and run multiple checks, which is fairly tedious. It's an inconvenience if you need to check a lot in a short window.

Results are reference-only

Like every other AI detector, Zhuque AI can't be perfectly accurate and can still get it wrong, both missing AI content and flagging human content by mistake. Tencent itself recommends combining it with human judgment. So don't use the Zhuque AI number alone to reach a high-stakes final conclusion.

The bottom line

Overall, Zhuque AI is a formidable free "scanner" for digital content at a time when AI content is increasingly everywhere and hard to tell apart by eye. Its brightest spot is high-accuracy detection of AI-made images and video, plus bundling all three formats onto one free platform. It's a sensible pick for teachers, editors, or content creators who need a quick verification layer before trusting or publishing a piece of material.

That said, keep Zhuque AI's "strong zone" clearly in mind: images, video, and Chinese/English text, not text in other languages. Treat the probability score the tool gives you as a reference signal to look more closely, rather than a final verdict. If you need a fast way to sanity-check how "real" an image or video is amid the AI content storm, Zhuque AI is well worth trying today.