I Let an AI Essay Writer Do My College Homework for a Week — Here's What Happened
Disclaimer: This article describes an approved experiment conducted for a journalism course on technology ethics, with full disclosure to professors. We do not endorse submitting AI-generated work as your own without proper disclosure, which may violate academic integrity policies at your institution.
When my journalism professor assigned our class a feature article exploring "how technology is changing traditional practices," I pitched an immersive experiment: I would use AI essay writers for my homework across different courses for one week and document the results.
With my professors' approval (and considerable curiosity), I embarked on this educational experiment. I would use leading AI writing tools to help with all written assignments across my five courses, documenting the process, results, and what I learned along the way.
What followed was a revealing journey into the capabilities and limitations of artificial intelligence in education—complete with surprising successes, notable failures, ethical dilemmas, and important revelations about the future of writing in the digital age.
Setting Up the Experiment
The Ground Rules
- All professors were informed of the experiment and could provide specific guidelines about how AI could be used
- Each assignment would note that AI tools were used and explain the process
- For research-based assignments, I would verify all facts and sources
- I would document my entire process, including prompts, editing time, and challenges
- Assignments would be submitted normally and graded according to standard criteria
I selected three different AI writing tools for my experiment:
GPT-4
The most advanced general-purpose AI model, using its ability to understand complex instructions.
EssayPro AI
A specialized academic writing tool with discipline-specific features for essays and papers.
ResearchBuddy
A research-focused AI with citation capabilities and access to academic databases.
My course load for the week included assignments from:
- Introduction to Political Science (analytical essay)
- Modern American Literature (critical response paper)
- Environmental Science (research-based report)
- Business Communication (professional email and proposal)
- Media Studies (content analysis)
Day 1: The Political Science Essay Challenge
The Assignment
"Write a 1,200-word essay analyzing how social media has impacted political polarization in the United States since 2016. Include at least three academic sources and address counterarguments to your position."
My Process
I started with GPT-4, providing it with the assignment prompt and asking it to generate an outline first. The outline it created was logical and comprehensive, with clear sections addressing:
- Introduction and thesis on polarization
- Echo chambers and algorithmic reinforcement
- Case studies (2016 election, COVID-19 responses)
- Quantitative evidence of increasing polarization
- Counterarguments (pre-existing polarization trends)
- Rebuttals with evidence
- Alternative factors beyond social media
- Conclusion and policy implications
After approving the outline, I asked GPT-4 to draft the full essay. It produced a well-structured piece in about 30 seconds. The essay had clear topic sentences, logical flow, and appropriate academic tone. However, I immediately noticed issues with the citations—GPT-4 had invented sources that didn't exist and attributed quotes to researchers who had never said those things.
I switched to ResearchBuddy for the citation problem, asking it to suggest real academic sources for each main point. This worked better, providing actual papers from journals like Political Communication and New Media & Society. I spent about an hour verifying these sources, reading abstracts, and integrating real citations into the AI-generated text.
Time breakdown: 5 minutes to generate the outline, 30 seconds for initial draft, 1 hour verifying and fixing citations, 30 minutes editing and personalizing. Total: About 2 hours (compared to the 5-6 hours I'd typically spend).
The Result
My professor gave the essay a B+ (87%), with comments that included:
"Well-structured analysis with good use of sources. The thesis is clear, though some of your analysis feels somewhat generic. I appreciate your disclosure about using AI tools, and I can see where you've added your own analysis. More original insights and specific examples would strengthen this into A territory."
My takeaway: The AI provided a solid foundation and saved significant time, but the essay lacked the unique perspectives and specific examples that come from deep engagement with the material. The fictional citations were a serious problem that required significant human verification.
Day 2: Literature Response Paper — Where AI Stumbled
The Assignment
"Write a 750-word response to Toni Morrison's 'Beloved,' analyzing how Morrison uses the supernatural elements to explore the psychological impact of slavery. Reference at least three specific passages from the text."
My Process
This assignment revealed the first major limitation of AI writing tools. I provided EssayPro AI with the prompt and asked it to generate a response. What I received was superficially impressive but fundamentally flawed:
The Problems:
- The AI referenced scenes that don't exist in the book
- It misinterpreted key supernatural elements, confusing characters and events
- Quoted passages were completely fabricated or grossly misquoted
- The analysis was generic and could apply to almost any book with supernatural elements
I tried again with GPT-4, providing it with specific passages from the book to reference. The results improved slightly, but the AI still struggled with the nuanced literary analysis the assignment required. It couldn't capture the complex symbolism or the deeper themes embedded in Morrison's writing style.
After two failed attempts, I realized I needed to write this one mostly myself. I used the AI only to help organize my thoughts and polish my prose. I wrote about 80% of the content, using GPT-4 to help refine my language and organization.
Time breakdown: 40 minutes on failed AI attempts, 2 hours writing my own analysis, 30 minutes using AI to help with editing. Total: About 3 hours (similar to my usual time).
The Result
My professor gave the paper an A- (92%), commenting:
"Thoughtful analysis with excellent close reading of the text. Your interpretation of how the ghost of Beloved embodies collective trauma is particularly insightful. I see from your disclosure that you used AI primarily for organization and editing—this feels like your authentic voice and understanding of the text."
My takeaway: AI tools struggle significantly with nuanced literary analysis and close reading of texts. They can invent content that sounds plausible but is factually wrong. For literature assignments requiring deep textual engagement, AI was more useful as an editing assistant than a content generator.
Day 3: Environmental Science Report — A Middle Ground
The Assignment
"Create a 1,000-word research report on a specific consequence of climate change in your local region. Include current data, projections, and at least one adaptation strategy. Minimum of five credible sources required."
My Process
For this assignment, I tried ResearchBuddy, which specializes in research-based content. I started by asking it to locate recent research about climate change impacts in the Pacific Northwest (my region). This is where the tool proved valuable—it could access databases and provide actual research findings with proper citations.
ResearchBuddy provided real data on increasing wildfire frequency, changing precipitation patterns, and impacts on salmon populations—all with proper citations to government reports and academic studies. I verified these sources and found them to be accurate and relevant.
I then asked it to generate a structured report incorporating this research. The draft it produced was scientifically sound but somewhat dry and technical. I spent about an hour revising to add a more engaging introduction about my personal observations of changing fire seasons, and to adapt the conclusion to include more specific local adaptation strategies I had learned about in class.
Time breakdown: 30 minutes for research gathering, 1 minute for initial draft generation, 40 minutes verifying sources and data, 1 hour personalizing and enhancing. Total: About 2.5 hours (compared to 4-5 hours normally).
The Result
My professor gave the report an A (95%), with the comment:
"Excellent research with well-integrated data and strong scientific backing. Your personal observations in the introduction effectively humanize the topic, and your adaptation strategies show good understanding of local context. Interesting to see your note about AI assistance—the tool seems to have helped with research gathering while allowing your understanding to shine through."
My takeaway: For fact-based, research-heavy assignments, specialized AI tools with access to academic databases proved extremely valuable. The key was using AI for what it does best (gathering and organizing information) while adding my own experiences and insights to make the content more engaging and personalized.
Days 4-5: Business Communication and Media Analysis — Mixed Results
Business Communication Project
The business communication assignment (writing a professional email sequence and project proposal) was where AI truly excelled. GPT-4 produced clear, professional business writing that needed minimal editing. My professor (who specifically approved using AI for this task) gave it an A and noted that business communication is increasingly AI-assisted in the professional world anyway.
Media Studies Analysis
The media studies assignment required analyzing representation patterns in a streaming show. This yielded mixed results—AI tools could help organize my observations and provide context about media theory, but couldn't actually watch and analyze the show's content. I had to provide all the specific examples and observations, making the AI more of a writing assistant than a content creator.
This assignment received a B, with my professor noting that the analysis had "good framework but needs more specific, nuanced examples."
The Final Tally: What I Learned
Grades and Time Investment
| Assignment | AI Contribution | Time Saved | Grade |
|---|---|---|---|
| Political Science Essay | 70% AI, heavily edited | ~4 hours | B+ (87%) |
| Literature Response | 20% AI, mostly editing help | ~30 minutes | A- (92%) |
| Environmental Report | 60% AI, research help | ~2 hours | A (95%) |
| Business Communication | 80% AI, minimal edits | ~3 hours | A (94%) |
| Media Analysis | 40% AI, framework only | ~1 hour | B (85%) |
Key Discoveries
What AI Did Well
- Generated clear, well-structured frameworks for almost any assignment
- Excelled at formulaic writing (business communications, reports)
- Provided research assistance and information gathering
- Helped overcome writer's block and starting anxiety
- Saved significant time on drafting and organization
Where AI Failed
- Often generated false or fabricated information with confidence
- Struggled with deep textual analysis and nuanced interpretation
- Created generic content lacking personal insight and original thinking
- Required significant fact-checking, especially for citations
- Could not incorporate personal experiences or observations
The Ethics Question
Throughout this experiment, I was constantly aware of the ethical dimensions of using AI for academic work. My disclosures to professors enabled open discussion about where the line between assistance and academic dishonesty might lie.
The consensus from my professors: Using AI as a brainstorming partner, research assistant, or editing tool was generally acceptable, especially with disclosure. Submitting AI-generated work with minimal human input or critical engagement was not. The key ethical factor was whether I was using AI to enhance my learning or to avoid it entirely.
Professor Feedback
"This experiment reveals the complex reality of AI in education—it's neither a magical solution nor a simple threat to learning. What matters is how students use these tools. Are they using AI to sidestep critical thinking, or as an assistant that helps them express their own understanding more effectively? The former undermines education; the latter may actually enhance it." — Dr. Rivera, Media Studies
Conclusion: The Future of Student Writing in the AI Era
After my week-long experiment, I've come to see AI writing tools not as either saviors or villains in education, but as powerful tools that require thoughtful integration. Like calculators in mathematics, these technologies can either enhance learning or shortcut it, depending on how they're used.
The most successful approach I found was treating AI as a collaborative assistant rather than a replacement for my own thinking. Using AI to help organize thoughts, provide research leads, assist with outlining, and polish prose yielded the best results, especially when combined with my own critical analysis, personal insights, and verification.
The tools saved me approximately 10 hours of work across a week of assignments—significant time that could be redirected to deeper reading, research, or thinking about the material. However, the assignments where I contributed the most original thought consistently received the highest grades.
As one professor noted during our post-experiment discussion: "The future belongs to students who can use AI effectively while maintaining their own intellectual voice. The skill isn't avoiding these tools; it's knowing how to collaborate with them productively while still developing your own abilities."
This balance—using AI to handle routine aspects of writing while focusing human effort on the elements that require genuine understanding, critical thinking, and personal perspective—seems to be the most productive path forward as we navigate education in the age of artificial intelligence.
About This Experiment
This journalistic experiment was conducted with full disclosure to professors as part of a technology ethics course. The purpose was educational exploration of AI's capabilities and limitations in academic settings, not to promote academic dishonesty. All universities and AI tools have different policies—always check your institution's guidelines before using AI assistance for coursework.
Other Articles You Might Like
Time Management for Students: Maximizing Writing Efficiency with Yomu
Yomu real-time feedback, grammar checks, outlining tools, editing features, vocabulary enhancement, collaboration capabilities, and multilingual support can help college students maximize writing efficiency, conquer the time crunch, and complete high-quality assignments without sacrificing personal time.
How to Ethically Use AI Essay Tools While Maintaining Academic Integrity
A comprehensive guide to incorporating AI writing assistance into your academic work ethically, with practical strategies for proper attribution, transparent use, and maintaining your authentic voice and intellectual contribution.
Do College Admission Officers Use AI to Detect AI-Written Essays?
An in-depth investigation into how college admissions offices are responding to AI-generated essays, what detection technologies they're using, and how students can ensure their applications remain authentic.
Can AI Writing Tools Teach You to Write Like Famous Authors?
Exploring how artificial intelligence can analyze and mimic the distinctive styles of literary legends—from Hemingway to Austen—and whether these tools can actually help improve your own writing through stylistic imitation.
Can an AI Paper Writer Handle APA, MLA, and Chicago Styles Correctly?
A detailed analysis of how well modern AI writing tools manage academic citation standards, with testing results across major citation styles, common error patterns, and practical strategies for students and researchers to get the most accurate citations when using AI assistance.
Could an AI Essay Writer Write a Viral Manifesto? We Tried It on Twitter
An experimental investigation into whether AI-generated manifestos can capture public imagination on social media. We tested how different AI systems perform when tasked with creating content designed to spread virally, and explored the ethical implications of algorithmically-optimized persuasion.