My Step 2 experience
After defending my PhD in April 2025, I had two months before clerkships and used this time to prepare for Step 2. I had fully matured the Anking Step 2 + Step 3 deck by mid-second year. I completed Amboss from April to May, averaging 70%. This identified gaps in obstetric care, trauma protocols, and quality improvement, which I addressed using Amboss articles. I finished UWorld from May to June, averaging 85%, likely because many Anking cards derive from UWorld. For incorrect questions, I reset and redid the corresponding Anking cards.
After clerkships began in late June, I completed a second Amboss pass by late July, focusing on weak areas (tremors, brain hemorrhage localization, immunodeficiencies) and using ChatGPT to generate comparison tables. In August and during my two-week dedicated period in September, I completed CMS forms 5-8, NBMEs 9-15, and both Free 120s. I created a PowerPoint of all incorrect answers and reviewed it three times. This phase focused on understanding how NBME asks questions rather than learning new content.
I preferred Amboss to UWorld. The questions were more nuanced and required closer attention to detail. Some say this causes overthinking, but I disagree. However, you must read all answer explanations, including incorrect choices, since Amboss lacks UWorld's textbook-style explanations.
| Exam | Date taken | Result |
|---|---|---|
| Free 120 (Mar 2023) | 06/22/2025 | 88% |
| NBME 9 | 07/27/2025 | 259 |
| NBME 10 | 07/29/2025 | 268 |
| NBME 11 | 08/01/2025 | 268 |
| NBME 12 | 08/22/2025 | 260 |
| NBME 13 | 08/25/2025 | 262 |
| NBME 14 | 08/27/2025 | 270 |
| NBME 15 | 08/28/2025 | 270 |
| Free 120 (Jul 2023) | 09/01/2025 | 90% |
My Amboss predicted score was 268 (95% CI: 259-277), with my actual score of 275 falling within the confidence interval.
I took the test on September 18, 2025. I finished each block in 40 minutes and spent 10 minutes reviewing flagged questions. I flagged 8-10 questions per block where I had moderate uncertainty. Most questions, including legal and ethics content, required straightforward reasoning. The more nuanced flagged questions remained unclear even after asking ChatGPT during breaks. The exam felt fair and well-represented by my preparation resources, though no single resource perfectly mirrored it. Thanks to Anking, all content felt familiar. I minimized re-reviewing questions to conserve stamina—by the final block, fatigue was significant. I estimate I got around 88% correct for my score of 275, though the exact conversion depends on form difficulty.