Residency Match Process in the Age of AI: Has Anything Actually Changed?

For medical students, Match Day is the moment years of training, exams, and sleepless application cycles have all been building toward, an outcome that decides where the next chapter of their career begins.

The system at the center of this process, which is run by the National Resident Matching Program (NRMP), is relatively straightforward.¹ As the NRMP explains, the algorithm is based on the rank order lists that both applicants and programs submit. In short, it is all about optimizing outcomes based on these preferences, rather than scoring or evaluating the applications themselves.² While the matching algorithm has stayed pretty much the same, advancements in AI technology are starting to change how applications are prepared, reviewed, and managed.

1. Screening applicants

One of the major pressures facing residency programs today is the number of applications they receive, with some programs even receiving thousands of applications annually for only a handful of positions.³ And when application volumes grow beyond what a program can realistically manage, review quality may suffer, for example, using simplistic cutoffs such as examination scores to filter applicants. Data show that between 30% and 65% of applications across specialties are screened out before anyone conducts a meaningful individual review.⁴ For instance, without a holistic review, a candidate who initially struggled on a licensing exam may never receive a closer look despite having strong specialty-specific clinical and research experience.

The advantage of AI in this context is that it can enable more comprehensive reviews. Early research exploring the application of AI shows promise, wherein AI tools reportedly identified strong candidates who had been unintentionally passed over when human reviewers were overwhelmed or focused too narrowly on one aspect of an application.⁵

That said, the risks of using AI also deserve equal attention. Studies acknowledge bias as a serious concern; AI systems trained on historically biased data may reproduce those patterns, reinforcing existing disparities.⁵ Further research in this field is needed to assess the extent of bias and whether these AI tools treat all applicants equitably.

2. Evaluating application components

Letters of recommendation are central to the residency application process but evaluating them fairly and consistently can be challenging. Some studies have identified differences in the language and tone used in these letters, including variations linked to the gender or race of applicants and letter writers.⁶ AI tools analyzing language patterns can help identify subtle differences in wording or sentiment that are not always obvious during manual review.⁶

Maintaining consistency when evaluating written application materials is another barrier. Although structured scoring systems are often used, they still rely on human review and can be affected by factors such as time constraints and reviewer fatigue.⁶ Early research suggests that AI tools may help improve consistency in scoring. For example, in one study, ChatGPT showed more consistent scoring than human reviewers in some cases when evaluating application materials using a standardized system.⁶ However, greater consistency does not necessarily mean that these evaluations reflect what residency programs are actually looking for in candidates, which is why these decisions cannot be fully automated.

3. Writing assistance

Writing letters of recommendation is also time intensive for faculty, many of whom write several letters each year. Reports from the Association of American Medical Colleges (AAMC) describe how some faculty are beginning to use AI tools to help draft these letters.⁷ In these cases, AI is often used to turn notes or bullet points into structured drafts, which are then reviewed and edited by the writer. Faculty report that this can save time and, in some cases, improve the clarity and flow of their letters.⁷ At the same time, this practice raises important concerns about data privacy and the need to ensure that human judgment remains central to the process, particularly when handling sensitive applicant information and finalizing the content of the letter.⁷

4. Standardizing transcript review

One persistent challenge in residency selection is that there is no single academic yardstick that works across all applicants. For instance, the way medical schools report grades is far from uniform (some may award honors while other may use a simple pass/fail grading system), meaning two applicants with similar academic profiles can look very different on paper simply because of where they trained.⁸

This is the problem Thalamus built its transcript normalization tool to address. By applying Large Language Model (LLM) to raw transcript data, the tool works to bring consistency to how grades and course names are read and interpreted across institutions, giving programs a more reliable basis for comparison.⁸

However, it should be noted that in some cases, for example, in the case of international scores, because of different grading systems, formats, and transcript styles, there isn’t enough reliable data to standardize grades accurately.⁸ Because of this, the system may only processes transcripts in which it is confident with the data, and may skip or flag others, meaning some students’ grades may not receive a assessments. On top of that, the AI system itself isn’t perfect and can sometimes misunderstand transcripts, so even Thalamus recommends its results should only be used as a support tool (not as the final decision maker) and programs should always double-check important details directly from the original documents.⁸

Conclusion

AI is beginning to reshape how residency applications are reviewed, with clear potential to make the process more efficient, and this shift is already underway in practice. The Association of American Medical Colleges (AAMC) has introduced Cortex, an AI- and machine learning–enabled platform developed by Thalamus, which has been available to all ERAS-participating residency and fellowship programs since July 2025.⁹ It is suggested that tools like this can substantially reduce the time required for application screening by around 50%, highlighting the practical impact AI may have on an already time-intensive process.⁹

At the same time, these benefits do not eliminate the underlying concerns. Recognizing this, the AAMC has outlined guidance for the responsible use of AI in residency selection, emphasizing the need to safeguard applicant data, reduce the risk of bias, and the importance of human insight in decision-making.¹⁰ As these tools become more widely adopted, their value will depend not just on efficiency gains, but also on how carefully they are implemented and monitored in practice.

Sources

1. 1. The National Resident Matching Program (NRMP). https://www.nrmp.org/
2. 2. The Matching Algorithm: How It Works. https://www.nrmp.org/intro-to-the-match/how-matching-algorithm-works/
3. 3. How Can AI Revolutionize the Match Day Process? https://www.facs.org/for-medical-professionals/news-publications/news-and-articles/bulletin/2025/february-2025-volume-110-issue-2/how-can-ai-revolutionize-the-match-day-process/
4. 4. The Use of Artificial Intelligence in Residency Application Evaluation—A Scoping Review. https://pmc.ncbi.nlm.nih.gov/articles/PMC12169010
5. 5. Artificial Intelligence in Residency Recruitment. https://www.neurology.org/doi/10.1212/NE9.0000000000200150
6. 6. Preparing for AI in Resident Selection: A Scoping Review of Current Applications and Limitations. https://onlinelibrary.wiley.com/doi/10.1002/lary.32308
7. 7. AI Will Now Draft Your Residency Recommendation Letter. https://www.aamc.org/news/ai-will-now-draft-your-residency-recommendation-letter
8. 8. Methodology for creation and processing of a novel Transcript Normalization Tool in Cortex Application Screening and Review Platform. https://www.thalamusgme.com/blogs/methodology-for-creation-and-processing-of-a-novel-transcript-normalization-tool-in-cortex-application-screening-and-review-platform
9. 9. AAMC-Thalamus Collaboration Information for Programs. https://www.aamc.org/services/eras-institutions/faqs-aamc-thalamus
10. 10. Principles for Responsible AI in Medical School and Residency Selection. https://www.aamc.org/about-us/mission-areas/medical-education/principles-ai