Graduate school recommendation letters are not college recommendations. Committees are not trying to figure out whether you are a good student in general. They are trying to figure out whether you can do independent research or advanced professional work at the level their program expects.
That changes who you should ask and how you should brief them. A specific, enthusiastic letter from a professor who supervised your thesis will do more than a generic letter from the department's most famous faculty member.
Who to ask, in order
- Research supervisors. Faculty who supervised your research directly. Nothing beats them.
- Professors who know you well. Usually from smaller seminars where you wrote real papers and participated.
- Thesis or capstone advisor. If you did one, they should be on your list.
- Work supervisors. Relevant for professional masters (MPP, MPH, MBA, and similar) and sometimes for PhD applicants with industry experience.
Who not to ask
- Famous professors who don't know you. The letter will read as generic and the committee will read it that way.
- Professors from large lectures. A B+ from a 300-person class does not produce a specific letter.
- TAs. Unless they directly supervised your research, this is not appropriate.
- Family friends or unrelated employers. Irrelevant to a research committee.
- Anyone likely to write a lukewarm letter. A tepid letter does active damage. If someone seems hesitant, thank them and move on.
How to actually ask
- Ask early. Six to eight weeks before the first deadline, ideally.
- Ask in person when you can. A short meeting signals seriousness and gives them room to say no.
- Ask for a strong letter, not just a letter. "Are you able to write a strong letter of support for my PhD applications?" gives them a graceful out.
- Explain your plans. Why graduate school, why now, which programs, what you want to work on.
- Follow up with materials. The packet below.
What to send your recommenders
- Updated CV. So they have the full picture.
- Statement of purpose draft. Even a rough one helps them align their letter with your pitch.
- Program list with deadlines. Clean, organized, with portal links.
- A reminder of your work together. Specific projects, papers, and what you contributed. Do not assume they remember.
- A short list of points you hope they can speak to. Research ability, specific skill, resilience, writing, whatever is true and useful.
What makes a strong letter
- Specific examples. A concrete project, not "great student".
- Context for comparison. "Top few percent of students I have taught in fifteen years" lands. "Very strong" does not.
- Evidence of research potential. Especially for PhD programs.
- Personal knowledge. A letter that could only have been written by someone who actually worked with you.
- Enthusiasm. "I recommend without hesitation" outperforms "I recommend".
Following up
- Two weeks out. A polite check-in if the portal still shows unsubmitted.
- One week out. Another short nudge if needed.
- After submission. A thank-you note, same day.
- After decisions. Let them know where you landed. They put real work in and want to know it paid off.
If you have been out of school for a while
- Reconnect with former professors. They often remember good students longer than you expect.
- Work supervisors are fair game. Especially for professional programs, and especially if they can speak to your analytical ability.
- Consider post-bacc coursework. One or two advanced classes can generate a fresh academic letter.
- Provide more context, not less. Assume they need the details refreshed.
The short version
- Pick recommenders who know you well over recommenders with famous names.
- Ask early, ask clearly, and give them a graceful out.
- Send a real packet, not just a thank-you.
- Follow up when you need to, politely.
Helpful resources
- Council of Graduate Schools - Guidance on the graduate application process
- NSF GRFP - Fellowship requirements and letter expectations for STEM applicants
- AAMC - Association of American Medical Colleges (for medical school applicants)