AI文本检测器

人工智能, 网络与网站, 文本相关

自由的
无需下载/安装
无需注册

You had manual writing tools, then recently AI automatic content writing tools, then plagiarism detectors … and now 人工智能检测人工智能编写的文本！

At least on the few trials we did …. indeed, it did detect humanly written texts and AI-generated tests.

文章没有说明它究竟是因为人类文本错误百出还是因为人工智能语法完美才将其识别出来！或许是所有参数共同作用的结果。

To be used by writers, anti-cheaters… or cheaters taking more precautions.

快速、免费、便捷。

请注意，这是为 GPT-2 开发的，而不是为最新版本 GPT-3 开发的。

学科： 教育、新闻、出版、学术研究、内容营销。

GPTZero、Turnitin、Originality.ai 的免费替代方案。

AI 文本检测器与 GPTZero、Turnitin 和 Originality.ai 的比较
AI文本检测器	GPTZero	Turnitin	Originality.ai
检测依据： A RoBERTa base classifier fine-tuned on outputs of the 1.5B-parameter GPT-2 model. It predicts whether a text came from GPT-2 and returns probabilities for real versus fake text.	检测依据： Scores pasted or uploaded text and highlights sentences flagged as likely AI-generated. The vendor names ChatGPT, Claude and Gemini among the covered models, and adds a plagiarism checker and an authorship replay feature.	检测依据： Reports an overall percentage and highlights suspected segments inside the Similarity Report. It also flags text that may have been reworded by paraphrasing tools, a check available in English only.	检测依据： Offers AI detection on every plan, including the free tier. Paid plans add plagiarism checking, full-site and URL scans and scan history.
Limits and validity: The developers report approximately 95% accuracy on text from the 1.5B GPT-2 model and say this is not enough for standalone detection. The model card warns against using it as a ChatGPT detector, and larger generators are harder to detect.	Limits and validity: Accuracy figures are the vendor own claims, and the vendor notes that no AI detector is perfectly accurate. Full language support is stated for English, German, Portuguese, French and Spanish.	Limits and validity: Needs at least 300 words of prose, at most 30,000 words and under 100 MB per file. The vendor aims for a false positive rate under 1% on documents with more than 20% AI writing, and scores below 20% appear only as an asterisk.	Limits and validity: The free tier allows 3 scans per day and 2,000 words per scan, with file upload of docx, pdf and doc.
许可证和费用： The model is published on Hugging Face under the MIT licence, so the weights can be downloaded and inspected.	许可证和费用： A free scan works without an account, and scans above 10,000 characters need a free account. The team plan lists an allowance of 500,000 words each month per member with a minimum of 2 seats.	许可证和费用： Licensed to institutions, as part of Turnitin Originality or as an add-on for iThenticate 2.0. Only instructors and administrators see the indicator.	许可证和费用： Pro and Enterprise are paid subscriptions, billed monthly. One credit equals 100 words, and the vendor page shows some inconsistent figures, so check current prices.
集成和 API： No API is documented on the model card. The model is intended for research on synthetic text generation and detection.	集成和 API： An API with ready-to-use code in 17 programming languages is offered, with a Chrome extension and integrations such as Google Classroom, Canvas and Zapier.	集成和 API： LTI integrations and a RESTful API are listed for Turnitin Originality.	集成和 API： API access is listed under the Enterprise plan, and a Chrome extension is included on paid plans.
预期用途： The model card excludes serious academic misconduct allegations from the intended use. It suits studies of how GPT-2 era detectors behave.	预期用途： Stated users include educators, students, publishers, recruiters and marketers.	预期用途： Built for educational institutions. The vendor states the score should not be the sole basis for action and does not determine misconduct.

涵盖的主题： AI Text Detector, plagiarism detection, content generation, automatic writing tools, grammar analysis, text analysis, human-written texts, AI-generated texts, detection algorithms, machine learning, natural language processing, accuracy assessment, anti-cheating tools, user-friendly interface, performance evaluation, GPT-2, and GPT-3..

Privacy: we have no feedback of potential problem with this tool and author. But like many online services, you are here providing them with your own data. Depending of the criticity of your data, this is your responsibility to assess author's reputation and potential misusage, and to check and accept their terms and conditions, or not.