Letters of Recommendation (LOR) are ubiquitous in academia. They are an important component of applications for schooling, career advancement, enrichment activities, and for awards. A good LOR can make an otherwise borderline application successful, and a bad LOR can destroy the chances of an otherwise good applicant. Faculty members, particularly those who teach often, can find themselves inundated with requests for LORs and can find the process overwhelming. However there are strategies that can make the process of writing and submitting LOR more manageable and help ensure your letter is interpreted the way you intend..

Note: Letters of Recommendation (LOR) are distinct from NIH Letters of Support (LOS) and from NIH Reference Letters for fellowships. This article has helpful information on how to request a LOR.

Deciding to Write a LOR

After receiving a request for a LOR from a student or colleague, you should first decide if you are both willing and able to write a strong LOR for the requestor. If you do not think that you can write a strong letter for the requestor, you should decline. Consider both your relationship with and opinion of the requester, as well as your available time when you make this decision. If you decide not to write the LOR, notify the requester as soon as possible so that they have a chance to find someone else who can write for them. Your declination can be a very simple statement such as “Dear [NAME], I’m unable to write a LOR for you at this time. Best of luck with your application.” There’s no need to provide the details of your overwhelming schedule, or if you don’t feel comfortable putting your name behind someone, enumerating the reasons why you don’t think they are a good applicant. In the rare case that they write back and protest, you are allowed to simply not respond to requests when you’ve already provided a clear answer.

Once you’ve decided to write the LOR, you should reach out to the requester for information, and provide a deadline that they need to send your requested info to you. At a minimum you should request 1) what opportunity the LOR is for, 2) instructions on how and when to submit the letter, 3) their CV, and 4) any critical information that the responder would like you to include in your LOR. If the opportunity has a document providing guidance for letter writers, ask for that. You can also request additional information to help you. Many faculty ask requestors to provide an initial draft letter. For complex application situations, like applying to graduate schools or faculty positions, it can be helpful to ask the requestor to provide you with a spreadsheet listing the schools and programs they are applying to, deadlines, and links to submit the LOR. For special enrichment programs, you can ask for information on the requestors experience and motivation for the program. Gather all this information in a single location (I like to file LORs in a single LOR folder with subfolders for each requestor specified Last, First) and set up calendar reminders for key deadlines. 

Components of a good LOR

A good letter of recommendation is tailored to both the requestor and to the opportunity that they are applying to. It should be on letterhead and signed. For situations like graduate school applications, you can generate a generic letter for the requestor, but you should update key information in the LOR (the University, Degree Program, etc.) for each opportunity that they are applying to.

The LOR should be addressed to the most specific group you are able to identify. Occasionally this information is provided in the request for the LOR, or in program guidelines. If not, do your best. “Dear Admissions Committee” is much better than “To Whom it May Concern”. 

Next you should clearly state your recommendation in unambiguous language. For example, “I am writing to provide my highest recommendation for admission of NAME to the NAME OF PROGRAM at UNIVERSITY.” You can then briefly state, in a sentence or two, why you are making this recommendation. Focus on the qualifications of the requestor, their fit with the opportunity, and, if applicable, on your knowledge of the opportunity. Frame statements to emphasize how the applicant would benefit the program. Something to the effect of “NAME is an inquisitive young scientist with strong research skills and a genuine passion for learning. Her interests in TOPIC align well with your program’s focus on RELATED TOPIC.”

In the next paragraph, you should explain who you are and your relationship to the requestor. Start with your position and affiliation, and details of your relationship with the applicant. For example, “I am an Assistant Professor in the DEPARTMENT at UNIVERSITY. REQUESTER spent 10 weeks in my lab as part of a summer research program during the summer of 2024. During this time we had almost daily contact, and I was able to witness firsthand how REQUESTOR approached research, and her excitement for learning.” Next, provide your relevant experience and expertise. Feel free to talk yourself up a bit. You want to come off as someone with the knowledge to discern a good applicant from a bad one.

In the following paragraphs of the LOR focus on the requestor’s qualifications for the opportunity. Look carefully at the opportunity they are applying to and write down key qualifications to comment upon in your letter. Some opportunities will provide these. Bold them in your LOR so the review committee can find the relevant section easily. Some examples are “Research Expertise”, “Academic Performance”, “Leadership”, “Teaching and Mentoring”. Address each of these sections in turn, providing as much detail as you can. For example, don’t just say that someone is an excellent researcher, include outcomes that document this like manuscripts, posters, and awards. Provide context to explain gaps in performance. Quantify things whenever possible: “REQUESTOR was one of only 5 students in a class of 52 to earn an A”, or “REQUESTOR was an exceptionally productive graduate student, publishing 5 papers in 5 years, 2 more papers than most students I have mentored typically produce.”

Finally, conclude the letter much like you opened it, restating your recommendation and the key reasons supporting it. Offer to respond to any follow-up questions, and then sign the letter. PDF the letter, submit, and then let the requester know that you have submitted the letter.

Using AI to draft LORs

Given the formulaic nature of many Letters of Recommendation and the ample source materials available, it is tempting to use a Large Language Model (LLM), commonly called a Chatbot or AI, to generate draft letters. This can be a timesaving strategy for generating draft letters that you subsequently edit. Below I include some strategies and context if you are interested in this approach to writing LORs.

First, you should make sure that you are not inadvertently sharing private information via the LLM. Many universities have secure Chatbots that you can use. Even with this, you should check the Chatbot settings and turn off any options that allow the bot to “learn” from the information you input. You should also anonymize any information you put into the bot, removing names, ID numbers, and anything else that could potentially be used to identify someone. Most universities have guidance on how to set up Chatbots.

Second, keep in mind that these models are probabilistic. They cannot learn, they do not think, and they don’t know if something is true or not. If they lack information to share, they will fabricate something. They cannot go to a webpage and extract information from it, or understand graphics. If you want it to use information on a specific webpage, you should pdf the webpage and then upload that to the Chatbot. Also, Chatbots aren’t worth fighting with if they get something wrong, as they won’t learn from the experience. If the AI takes the letter in an odd direction, don’t waste time trying to correct it, just start a new session.

Third, be very specific in what you want the Chatbot to do. Upload all the source documents you have, and tell the Chatbot how you want it to use the documents. Tell it who you are and what specifically you are doing. If you want it to use specific subheadings for sections of the LOR, tell it what those should be and what should be included. 

Once a draft is generated, edit it. First, make sure all information included in the LOR matches information in one of your source documents. Verify citations of papers, GPAs, and award details. Read technical information carefully and make sure it is an accurate reflection of the requestor’s contributions. Also edit for style and tone so that your LOR sounds like you. Make sure everything is the right length and level of detail. You can ask the Chatbot to expand or contract different sections of the LOR. I’ve had the best luck when I’ve isolated the section of the LOR and provided specific guidance on the type of edits I want, for example, “shorten the following text by 20% while preserving tone and meaning”.

Conclusion

Writing letters of recommendation doesn’t have to be stressful. With some organization up front and a clear plan for requesting, drafting, and editing, the process can be relatively efficient and provide a means for you to make a positive impact for other scientists.