About the Plagiar-Ezy Project

The Plagiar-Ezy project represents a comprehensive academic research initiative designed to systematically evaluate the current capabilities of Large Language Models (LLMs) in producing university-level academic writing that meets institutional assessment standards.

Research Objectives

Primary Objective

To quantitatively and qualitatively assess how effectively current AI systems can produce academic work that meets university assessment standards across multiple disciplines and academic levels.

Secondary Objectives

Document the time efficiency of AI-assisted academic writing compared to traditional student work
Categorize levels of human intervention required for AI-generated academic content
Provide empirical evidence for discussions about AI’s impact on educational assessment
Inform the development of AI-resistant evaluation strategies
Contribute to policy discussions about academic integrity in the age of artificial intelligence

Methodology

The Test Framework

The Plagiar-Ezy test involves a systematic simulation of university assessment scenarios where AI systems are tasked with producing academic work equivalent to what would be expected from students. Our methodology includes:

Test Parameters

Scope: Multiple university modules across various disciplines
Assessment Types: Essays, reports, analysis papers, and other written assignments
AI Role: Acting as a sophisticated academic writing assistant
Human Role: Minimal engagement, simulating various levels of student involvement
Output Goal: Work that would achieve strong passing grades in university settings

Evaluation Metrics

Academic Quality: Assessed using standard university marking criteria
Time Efficiency: Measured from task initiation to submission-ready work
Human Intervention: Categorized levels of human input required
Disciplinary Competence: Evaluation across different academic fields
Assessment Vulnerability: Analysis of which assessment types are most susceptible to AI assistance

Intervention Categories

We classify human involvement in AI-assisted academic work across several categories:

Minimal Intervention: Basic prompting and formatting, representing students who simply request AI to complete assignments

Guided Intervention: Strategic prompting with some content direction, representing students who understand how to effectively collaborate with AI

Collaborative Intervention: Substantial human input in planning, reviewing, and refining AI outputs, representing sophisticated AI-human partnerships

Expert Intervention: Extensive human expertise applied to AI outputs, representing the upper bound of AI-assisted academic work

Academic Context

The Challenge

The rapid advancement of Large Language Models presents unprecedented challenges to traditional academic assessment methods. Current AI systems demonstrate remarkable capabilities in synthesizing complex academic arguments, applying disciplinary terminology, and following academic writing conventions.

Research Gap

While anecdotal evidence of AI-generated academic work exists, there has been insufficient systematic research examining consistent performance across disciplines, variation in effectiveness across assessment types, and the role of human intervention in optimizing AI outputs.

Policy Implications

Our research directly informs several critical policy areas:

Assessment design and AI-resistant evaluation methods
Academic integrity policies for the AI era
Pedagogical adaptation for AI-enhanced education
Institutional responses to AI tools in academic work

Ethical Framework

Research Ethics

This research operates under strict ethical guidelines:

No Actual Submission Fraud: All test submissions are conducted with appropriate disclosure or in controlled environments
Educational Purpose: Research aims to improve rather than undermine educational systems
Transparency: Full disclosure of AI involvement in all research outputs
Responsible Disclosure: Results shared to inform rather than enable academic misconduct

Academic Integrity Commitment

While this research explores AI capabilities in academic writing, we strongly support:

Institutional academic integrity policies
Transparent use of AI tools in education
Development of AI-adapted assessment methods
Continued value of human learning and development

Scope and Limitations

Project Scope

Temporal: Testing conducted during 2024-2025 academic year
Geographical: Focused on UK higher education standards and practices
Technological: Primarily using Claude 4 family models, with comparative analysis of other leading LLMs
Disciplinary: Spanning humanities, social sciences, and interdisciplinary studies
Academic Levels: Undergraduate and graduate coursework

Acknowledged Limitations

Sample Size: Limited to modules accessible to research team
AI Evolution: Rapid development means findings reflect specific temporal capabilities
Assessment Variability: Different marking standards across institutions and assessors
Ethical Constraints: Research conducted within boundaries of academic integrity

Expected Impact

Immediate Outcomes

Comprehensive dataset of AI performance in academic contexts
Empirical evidence for policy discussions
Identification of vulnerable assessment types
Documentation of best practices for AI-resistant evaluation

Long-term Impact

Contribution to educational methodology reform
Enhanced understanding of human-AI collaboration in academic contexts
Improved assessment design across higher education
More nuanced policies regarding AI use in academic work

Transparency Declaration

This website, including this content, has been developed using Large Language Models as part of our research methodology. This demonstrates the current capabilities that our research seeks to understand and address in academic contexts.

For more detailed information about our methodology, please see our [research documentation] or contact us directly.