Ensuring Academic Integrity in the Age of AI is a complex and urgent priority for all institutions–online, hybrid, and even face-to-face–due to the deep integration of learning to learning management systems.
Valid Assessment – An Institutional Priority
Educators and administrators must know what is being submitted, how it was created, and who was the learner. Securing the learning management system is essential to valid assessment. Today, the advancements in Artificial Intelligence systems designed for computer use–proficient at website navigation–are a threat to the value of all learning management systems and courseware environments. Despite the risks of AI on the LMS, including Agentic browsers, Computer-use Agents, Browser-based AI, several leading LMSes suggested that AI agents are not detectable.
This is a lie. Cursive detects AI agents.
Meeting the Challenge
At Cursive, we know that an individual can be verified through their unique behavior and computer interaction (TypeIDTM). We’ve extended this insight and approach to detect Agentic AI (and other tools) across the LMS. Our complementary text editor, local, and optional authentication plugin architecture create a comprehensive behavioral signals and interpretation system that can be used to reliably detect several forms of Agentic AI, student-AI tool use, and human impersonation. This includes, but is not limited to: traditional contract cheating, Perplexity Comet browser, Opera Neon browser, Companion AI, OpenClaw, Claude for Chrome, Google Lens.
Our continuous research and development to validate learning has led to our release of the Agent Detection Service: a comprehensive, multi-layered signal interpretation system.
Operational Impact & Efficiency
Cursive’s lightweight, customizable, and efficient approach provides a robust approach to identify non-human and impersonation behavior for logged in users across the LMS (quizzes, assignments, lessons, etc.). Comprehensively, this automatically logs and interprets 1000s of new signals making interpretation available to administrators in real time or faculty and evaluators after the fact.
Despite the high-fidelity, our approach minimizes server-side impact, resulting in an overall load that is less than the typical quiz load. At Cursive we strive to prioritize integrity without compromising the user experience or system performance.
The Cursive Detection Modules
| Module | Signal Count | What It Covers |
| Writing | 1,000s+ | Typing behavior, cut, copy, paste, revisions, duration, and input type. |
| Interaction | 27 | Mouse, click, keystroke, scroll, and comet-specific behavioral patterns. |
| Fingerprint | 32+ | WebDriver, headless indicators, automation globals, canvas/WebGL, extension probes, and network detection. |
| Injection Monitor | 30+ | DOM mutation detection for AI extensions, shadow DOM, overlay elements, and text patterns. |
| Request Analyzer | 20 | Server-side HTTP header analysis (UA, Client Hints, Sec-Fetch, Accept, consistency). |
Trailblazers in Academic Integrity
Our work on impersonation prevention, continuous authentication, agent detection, and addressing over-the-horizon challenges is based on years of experience developing Cursive and working with partners like Augsburg University, Shanghai International Studies University, and Murdoch University to support their valid assessment initiatives.
We have established a strong R&D pipeline through our Certified Integration Partnership with Moodle, pushing the envelope on detection capabilities. Additionally, our team is actively engaged in strategic conversations with LMS leadership to address this critical need across the market. Our ongoing investment in research is what keeps our solution ahead of the curve in a constantly evolving landscape, future-proofing your institution’s assessments.
Free – Agent Detect Lite
Our Agentic Browser detection tool is available at Moodle.org and provides a base, free level of Agentic detection: https://moodle.org/plugins/local_agentdetect