Question quality
Every answer, double-checked.
An AI model (Claude Haiku) solves each practice question on its own, without seeing our answer. When its answer differs from ours, the question goes to a person for review before students rely on it.
13,917
of 30,093 questions checked so far
91.9%
agree with our answer key
| Exam | Checked | Agreement | Sent for review |
|---|---|---|---|
| SAT | 5,917 | 94.3% | 340 |
| ACT | 8,000 | 90.2% | 785 |
| MCAT | 0 | — | 0 |
Agreement means the AI reached the same answer we did. It isn't a guarantee that any question is perfect, and it doesn't predict anyone's score. Questions are never shown here, so no answers are given away.