Artificial intelligence (AI) has become an integral part of our daily lives, with its influence increasingly felt in the education sector.
As more and more students turn to AI-powered tools to write their assignments, teachers face a growing challenge: how can they detect whether a piece of work was written by AI or by the student themselves?
In this article, we will explore how teachers check for potential AI-generated content.
Summary:
Some call it AI detection, others call it AI checking. What is this technology?
AI detection, also known as AI checking, refers to the technology used to identify content that has been potentially generated by AI.
As AI tools become more advanced, particularly in natural language processing (NLP) and content generation, it’s increasingly difficult to tell whether a text was written by a machine or a person.
AI detection systems analyze various elements of a document, such as writing style, word choice, and patterns of language use, to determine whether the content shows characteristics typically associated with AI-generated writing.
Mainly using machine learning and through training on a large corpus of AI and human texts, AI detection systems became able to differentiate between AI and human writing styles.
They can thus help educators to spot inconsistencies or anomalies that may signal the use of ChatGPT, Gemini, Claude or other AI tools
A few months after ChatGPT's arrival in the public sphere (late 2022), the first AI checkers, including the Compilatio AI detector, were released.
AI detectors are often retrained to adapt to new AI models (e.g. GPT-4.5 or Gemini 2.0) and improve their reliability rate.
Teachers and professors verify the use of AI in student work for various reasons: uphold academic integrity, assess students' essential skills, ensure equity between students, etc.
AI-generated content can compromise the learning process, making it crucial to detect and regulate its use.
Here are a more details on the reasons behind AI detection:
Professors use AI detection tools to identify passages that may have been generated by artificial intelligence. AI detectors analyze writing patterns and flag suspicious sections that resemble the style of generative AI.
When such passages are detected, teachers can ask students questions about these passages to check their understanding and whether they have actually written them.
In addition to AI detectors, professors use other strategies to verify the authenticity of assignments:
Comparing multiple submissions: AI users may submit nearly identical responses with probably the same examples.
Teachers rely on AI detection tools to identify potentially AI-generated content, with solutions like Compilatio Magister+ standing out. This tool combines AI detection and plagiarism detection, allowing educators to check for similarities and AI-generated text in a single analysis.
Many institutions equip their entire faculty with institutional tools such as Compilatio or Turnitin, which offer key advantages:
Some teachers also turn to personal use tools or even plagiarism detection software for students. These solutions can help compare content against web sources but lack advanced features like LMS integration, which automates tasks and saves valuable time.
Generally, AI detectors provide indicators (flags), not definitive proof. Teachers must analyze flagged sections, ask follow-up questions, and assess whether the students have really understood the subject and have written the final report.
The reliability of AI detectors varies depending on the specific tool and its technological advancements.
Most AI detectors regularly update their technology to adapt to new versions of AI.
For instance, Compilatio's AI detection engine boasts an accuracy rate of 99%, meaning it correctly labels 99 out of 100 text passages as either human-written or AI-generated.
A fire alarm metaphor to help you understand this percentage:
with an accuracy rate of 99% for a fire alarm: 99% of situations (fire or not) are correctly valued.
Additionally, Compilatio's detection engine maintains a false positive rate below 1.5%. A false positive occurs when the system incorrectly identifies human-written text as AI-generated.
Let's take the fire alarm example again:
In the case of a false positive, the alarm sounds for no reason, e.g. due to cooking steam, which can be classified as fire smoke.
For example, with a 1.5% false positive rate for an AI detector, out of 1,000 human-written passages, approximately 15 might be mistakenly flagged as potentially AI-generated.
It's important to note that while high accuracy and low false positive rates are desirable, no AI detection system is infallible. Educators should use these tools as aids, complementing them with their judgment and additional verification methods, for example exploratory questions on the part concerned, to maintain academic integrity.
To reduce AI-assisted plagiarism, teachers can set clear guidelines on AI usage, design personalized assignments that require critical thinking, and use oral questioning or in-class writing tasks to assess students' true skills.
Educating students on academic integrity and the cons of AI also helps foster responsible usage.
Students should use AI as a learning aid rather than a replacement for their own work. Always verify facts, cite sources properly, and develop personal writing skills to avoid over-reliance on AI. Understanding the risks of misinformation and detection can help students make ethical academic choices.
As the use of AI becomes increasingly common in teaching, it becomes necessary to check the authenticity of texts with an AI detector, as is the case for plagiarism.
If you're a teacher, use plagiarism checker for teachers to ensure the authenticity of student work by detecting both plagiarism and potential AI-generated content.
If you're a student who has used AI, first check your institution's regulations to see if it's allowed. Then, verify your sources using a plagiarism software for student that includes AI detection. Don’t forget to cite your use of AI according to your institution's quoting standards!
You may also be interested in these articles:
This informative article, which does not require personal reflection, was partially written with the assistance of ChatGPT. The automatically generated content has been revised (including corrections for repetition, sentence structure, added details, added citations, and fact-checking).
Cite this article according to APA standards
