Of the many beneficial uses for AI, one of them is to help build a physician residency application. But in that role, AI (augmented intelligence) should remain a supporting cast member—not the star.
During a recent presentation at the AMA IMG Distinguish Yourself Summit, an AI and engineering expert outlined the advantages of AI—also called artificial intelligence—for residency applicants. It can save time and sharpen messaging in finalizing a residency application, said Ankita Bhatia, who recently concluded her tenure as the AMA’s director of AI strategy and engineering. She also highlighted where it can actually weaken an application.
What are some practical use cases for AI in developing an application? What are the red lines? Here’s one expert’s take.
The rules of the road
While the Association of American Medical Colleges (AAMC), which oversee the MyERAS platform through which most aspiring residents apply, permits AI use in certain areas of the application creation process, there is a clear line—AI cannot be an author.
What does that mean for AI use on your application? AAMC guidance on personal statements offers definable guardrails, stating “the use of AI tools is acceptable for brainstorming, proofreading or editing the personal statement, but the final submission should represent your own work.”
Speaking more broadly, Bhatia said AI isn’t equipped to tell your story.
“In 2026, everyone has access to all these AI tools,” she said. “To distinguish yourself, it [your story] needs to be ungeneratable. The things that make you unique, that make you special, AI cannot create.”
If you use AI practically, however, it can “help call out all those things accurately.”
Make AI a critic
For both ethical and quality reasons, AI should not play a role in the creation of any part of your application. But it can be a mechanism for feedback that leads to improvement.
Bhatia described impactful use cases for AI by asking large-language model tools to flag vague language, identify unsupported claims and pressure-test application material
“If you write first, you can say: Hey, AI, can you critique this for me? Can you tell me where I am missing A, B and C,’” Bhatia said. “That’s where the power starts shifting; where you can really utilize it to get a benefit out of it.”
For international medical graduates (IMGs), specifically, Bhatia touted AI tools as offering value in contextualizing. While the raw materials—the facts of an IMG’s clinical life—must be your own unique inputs, AI can help you in framing them.
“What you should be using AI for is helping close the friction gap, helping you express yourself better in English, navigate CV systems and U.S. conventions, and translate job titles from your home country into terms that would make sense to U.S. reviewers,” she said.
The AMA offers key Match guidance for IMGs, including best practices for residency interviews. Learn more about navigating the process of practicing medicine in the U.S. with the AMA IMG Toolkit (members only).
Prompts have a purpose
Bhatia said applicants should be deliberate in how they prompt AI. An effective prompt typically includes:
- Role—define exactly what the model should be, such as “copy editor, not a rewriter.”
- Context—provide the real material and facts yourself.
- Task—give the model one clearly defined job at a time.
- Constraints—spell out what the tool should not do.
- Output format—specify how you want the response returned.
What does this look like in practice? Here’s an example of how to have generative AI tools make edits to application sections.
“When you use AI, give it a specific role, like copy editor, asking it to focus on grammar, consistency and structure rather than changing your voice,” Bhatia said. “Provide the real context yourself, be clear about what it should not do—rewriting sentences—and ask for a list of edits so you can review every change.”
Lean on source material
While it may act as though it has all the answers, AI tools aren’t always reliable in offering the most up-to-date information. They may also create answers that are simply untrue.
Because of that, Bhatia cautioned any residency applicant that an AI chatbot is a poor source for deadlines, visa details, program requirements, score benchmarks and citations. Instead, she urged applicants to verify those details with primary sources such as FREIDA™, the AAMC’s Residency Explorer tool and residency program websites.
“A chatbot’s memory is not a source,” she said. “It can invent program details, sponsorship policies, deadlines and score cutoffs fluently and without hedging.”
FREIDA is a free-to-use application that allows you to search for a residency or fellowship from more than 13,000 programs—all accredited by the Accreditation Council for Graduate Medical Education (ACGME).
Give a human the last review
Bhatia repeatedly stressed that AI should never be the final reviewer of a residency application. Mentors, specialty-specific advisers and faculty members with experience in residency selection can catch red flags that an AI tool has glossed over.
“AI should never be the last readers,” Bhatia said. “You need a human in the loop on all your outputs ... AI cannot tell you what is truly you.”
She added that when applicants are writing a personal statement, CV or filling in the application’s experiences section, AI has value in editing. “But you want a mentor to read it. They can tell you if it sounds like you and ask questions and make sure you are providing enough information.”
