Everyone claims their AI tool writes "university-level" essays. We put six of them against a real McGill philosophy rubric (PHIL 240, 1,500 words, applied ethics question) plus a human Canadian writer as control. Here's what happened.
The setup
Results (blind-grade average)
| Tool | Grade | AI-detection | Notes |
|---|---|---|---|
| Human Canadian writer (TutorsGallery) | A- (82%) | 4% AI | Strong thesis, real citations, Canadian context |
| Claude Sonnet 4.5 | B (74%) | 81% AI | Coherent but generic, hallucinated 2 citations |
| GPT-5 (paid) | B- (72%) | 93% AI | Predictable structure, hallucinated 3 citations |
| Gemini 2.5 Pro | C+ (68%) | 88% AI | Repetitive, missed the applied-ethics angle |
| Perplexity Pro | C+ (67%) | 76% AI | Better sourcing, weak argument |
| EduBirdie AI | C (62%) | 95% AI | Templated, generic, hallucinated citations |
| ChatGPT (free) | C- (58%) | 96% AI | Surface-level, no real engagement with the question |
What the AI tools got wrong
What the human writer got right
The practical takeaway
For real Canadian-graded work, see our essay writing service Canada.
Need expert help?
Skip the stress — get matched with a Canadian expert.
15% off your first order. Delivery from 3 hours. 100% original.
Order Now