Provisional: 1 test run. Every score below comes from a single run of our own test. See how we test.
The test
Give tutor-style feedback on a fictional essay paragraph containing two factual errors, explain the key facts at Year 12 level, and write a 5-question quiz. A good tutor corrects the errors without ghost-writing.
Results
| Tool | Plan tested | Score |
|---|---|---|
| Perplexity | Free plan | 84/100 Provisional: 1 test run |
| Claude | Opus, paid plan | 79/100 Provisional: 1 test run |
| ChatGPT | "Instant" model | 73/100 Provisional: 1 test run |
Perplexity: 84/100 (Provisional: 1 test run)
Caught both errors, cited figures, and asked the student to rewrite the paragraph themselves. The quiz took about 90 seconds and only question 1 was readable.
| Criterion | Points |
|---|---|
| Factual accuracy | 30/30 |
| Pedagogy | 24/25 |
| Quiz | 8/20 |
| Clarity | 14/15 |
| Friction | 8/10 |
Claude: 79/100 (Provisional: 1 test run)
Caught both errors and guided the student, but also supplied a full rewrite. The quiz answer key did not match the option order.
| Criterion | Points |
|---|---|
| Factual accuracy | 30/30 |
| Pedagogy | 15/25 |
| Quiz | 10/20 |
| Clarity | 14/15 |
| Friction | 10/10 |
ChatGPT: 73/100 (Provisional: 1 test run)
Caught both errors but ghost-wrote a full paragraph. The quiz widget hid the answers and then disappeared.
| Criterion | Points |
|---|---|
| Factual accuracy | 30/30 |
| Pedagogy | 13/25 |
| Quiz | 8/20 |
| Clarity | 14/15 |
| Friction | 8/10 |
Caveats
The quizzes were only partly verifiable: each tool showed its quiz in an interactive widget, and we could only capture question 1, so quiz scores rest on limited evidence. We plan a second run of every tool before treating any ranking as settled.