ChatGPT Emails to Professors: What They Notice

ChatGPT Emails to Professors: What They Notice

An Indiana University faculty page says generic emails that praise a professor's work without showing any familiarity with it are "extremely unlikely" to get a response. A University of Chicago lab page, checked October 31, 2026, now says clearly AI-generated emails are "unlikely to receive a reply".

The question applicants search is whether a professor can "tell". That is the wrong question. The faculty pages below complain about what the email is missing: a named paper, a real connection, a sign that you did the reading. An AI draft leaves those out by default, because it does not know them. Rewording the draft does not add them. Only you can.

Every faculty and university page quoted below was checked on October 31, 2026. Faculty pages change without notice, and two of the three state no date. For when an email is worth sending at all, see emailing a professor before you apply to a PhD. For finding the right people, see contacting a PhD supervisor.

What faculty pages say about AI-drafted and generic emails

These are three faculty members' own pages, not a survey. Read them as what these readers say, not as what every professor thinks.

The full note on Blekhman's lab page reads:

"A note on AI: I read every email myself, and I am much more likely to respond to one that you wrote in your own voice and your own words. Emails that are clearly AI-generated are unlikely to receive a reply"

His October 4, 2026 post explains why. He says he now receives "20 such emails a day" and that "almost all of them are AI-generated". Two of his reasons matter most for your draft. First: "Every one of these emails looks exactly the same." Second, on trust: "If the email tells me that you’ve read our latest paper and found it relevant, I have to ask myself, “have you really”?"

Blanco names the giveaway words. His page says emails "will contain words like "respected" or "esteemed" when referring to someone's home institution or work, without actually including any indication that one is familiar with either." He adds: "This type of emails is counter-productive and it is extremely unlikely anyone will respond to them."

Evans's page reads like a list of what a chatbot adds by default. "Don't waste space and time telling me how brilliant I am." "Don't attach anything to your email." He is not against all warmth: "A touch of flattery never hurts, but don’t go overboard."

Our view: the tool is not the main failure. The email is. An email with nothing in it that only you could write fails with these readers whether a chatbot wrote it or you did. An AI draft simply produces that email every time. Swapping "esteemed" for "respected" changes nothing, because Blanco's complaint is about the missing familiarity, not the word.

The countercase: for a reader like Blekhman, effort is part of the signal. His first reason is "You did not spend time writing it, so I am not going to spend time reading it." So a specific email that still reads as generated may go unread in his inbox. That is why our recommendation is to write the email yourself, not to edit a generated draft until it looks specific.

The AI email, dissected

Constructed example: a typical AI-drafted email

Constructed example, written by us, not a real applicant's email. The names, the lab and the field are invented. It is 117 words, counting the subject line and sign-off.

Subject: Prospective PhD Student: Research Interest

Dear Esteemed Professor Menon,

I hope this email finds you well. I am writing to express my keen interest in joining your esteemed research group as a PhD student. Having delved into your groundbreaking work, I was deeply inspired by your innovative contributions to environmental data science, which align perfectly with my research interests.

I am a hardworking, passionate and highly motivated student with a strong foundation in data analysis. My academic journey has showcased my ability to tackle complex problems, and I am confident I would be a valuable addition to your team.

I have attached my CV, transcripts and research proposal for your kind consideration.

Best regards, Ananya Rao

A faculty reader can check nothing in it. Line by line:

"Delved" and "showcased" are also on a measured list. A study of more than 15 million PubMed abstracts from 2010 to 2024 found that "Less common words with strong excess usage included delves (r = 28.0), underscores (r = 13.8), and showcasing (r = 10.7)". Here r is how many times more often the word appeared in 2024 than expected. The same study found that "at least 13.5% of 2024 abstracts were processed with LLMs" (Kobak et al., Science Advances, July 2, 2025). That study measured biomedical abstracts, not emails. It shows these words rose sharply once such tools spread. It does not show that professors scan for them.

Deleting those words does not rescue the draft. Take out every adjective and the email says only: a student wants to join, has a background in data, and has attached files. That is the problem the faculty pages describe.

Constructed example: the same email rewritten

The structure is the four-line method from emailing a professor before you apply: one thing you read, what you did that connects to it, who you are and your entry cycle, and one question with a short answer.

Constructed example, written by us, not a real applicant's email. The paper, the project and the numbers are invented. It is 79 words, counting the subject line and sign-off.

Subject: Fall 2027 PhD applicant: question on your 2025 sensor-drift paper

Dear Professor Menon,

Your 2025 paper corrected drift in low-cost air-quality sensors using a reference station nearby. In my final-year project I calibrated 12 such sensors with no reference station, using co-located temperature data instead.

I am finishing a BTech in environmental engineering and applying to your department's PhD for Fall 2027. Do you expect to take a new student in that cycle?

Project write-up: example.edu/arao/drift

Ananya Rao

Every sentence is checkable, and no tool could have written it without the writer's own facts: the paper by year and topic, the project with a number and a method, a one-line question, and a plain-text link instead of an attachment, as Evans asks. The reader has no reason to ask Blekhman's "have you really".

Can you use AI for the email at all?

Keep three things apart: what is published, what we recommend, and what nobody has published.

What is published. Cambridge's rules for applicants cover application documents and interviews. Its postgraduate page, Using AI in your application, sets three rules:

"AI must not be used when writing personal statements and CVs. This prohibition covers the use of AI to help your use of English."

"Appropriate use of AI tools to support the preparation of research proposals is permitted. Content generated by AI should be acknowledged."

"AI must not be used to impersonate or assist you during an interview, including to help your use of English. At interview, you should expect to be asked to confirm that you are not using AI."

The page does not mention emails to potential supervisors. How universities word their rules for the proposal itself is in AI rules for a PhD research proposal.

What nobody has published. The university rule we checked sets nothing for emails sent before you apply. That silence is not permission. It means the faculty member's own page, like Blekhman's, is the rule that applies to their inbox.

Our recommendation. Write the email yourself. Blekhman draws his line here: "The line I am drawing is between using AI to improve and polish writing you have done, and using AI to avoid writing and thinking altogether." In a second language, short and plain beats polished. Then watch where the email's sentences go next. The pairing you wrote to the professor belongs in your statement of purpose, and at Cambridge a personal statement may not use AI even for English, so a tool-polished paragraph cannot be reused there. If the professor interviews you, you will discuss the paper you named with no tool to help.

Before you send: the only-you test

Run these on your draft. Each has a pass condition.

1. The paper. Without looking, can you name the paper (year and topic) and one result in it? Pass: yes. If not, read it before you write.
2. The only-you sentence. Is there one sentence about something you did, with a number, a method or an object in it? Pass: yes, and it connects to the paper in check 1.
3. Praise for them. Delete every adjective about the professor or the university ("esteemed", "groundbreaking", "renowned"). Pass: the email loses no information.
4. Praise for you. Delete every adjective about yourself ("hardworking", "passionate"). Pass: the email still shows what you can do.
5. The question. Does the email end with one question the professor can answer in a line? Pass: yes.
6. Attachments. Does the lab's own page ask for files? Pass: you attached only what it asks for. If it asks for nothing, send a plain-text link.
7. Your words. Read it aloud. Pass: you could say every sentence in an interview without notes.

Questions about AI and professor emails

Can professors tell an email was written with ChatGPT?

None of the three pages describes a tool. They describe reading. Blanco says mass emails are "very obvious to us faculty members". Blekhman says the AI emails all look "exactly the same". Whether a given professor guesses right about a given email is unknown. What is known: an email with no checkable detail reads the same either way.

Can I use AI to find which professors to email?

No page we checked sets a rule for this. The risk is the list itself. Blekhman's worry is exactly this case: "maybe an agent generated a list of labs, sent emails to all of them, and the human candidate is not even aware my lab is on the list." Whatever builds your list, you still have to read one paper per professor before you write. That step is check 1 above. Before you commit to a lab, see how to vet a PhD advisor for red flags.

Should I say I used AI to fix my English?

No page we checked asks for a declaration in an email. If the same text goes into a statement or proposal, that document's rules apply, as in the Cambridge rules quoted above.

Should I attach my CV?

Follow the lab's page, because these two disagree. Evans says "Don't attach anything to your email." Blekhman's page asks graduate students to email "providing a short introduction, a CV, your transcript, and contact information for two references." If the lab's page says nothing, send no files and give a plain-text link.

How long should I wait before following up?

Evans says: "If you don’t get a reply after about a week, send a follow up email that politely asks if the message was received and includes the previous message." The timing by route and three constructed follow-ups are in follow-up email to a professor with no reply. If the professor did reply, see replying to a professor's positive response or professor replied "apply through the program".

Get a free review of your statement of purpose

Get a free Quick Review of your statement of purpose against GradPilot's PhD adviser-fit rubric. It reads whether you name a research direction a faculty reader could match to a person or group, whether your reasons name what that group actually works on, and whether your claims about other people's work are specific and checkable. Those are the same things your email needed. For the fast first read a professor gives a statement, run the Professor's First Pass lens. More doctoral guides are in the PhD essays hub, and the PhD statement of purpose review page is the starting point for doctoral statements.

Sources

Retrieved October 31, 2026, except where noted.

• Ran Blekhman, University of Chicago, Join the Lab (no date stated): the note on AI-generated emails; what graduate students are asked to send.
• Ran Blekhman, Why I am not reading your AI-generated email (October 4, 2026): email volume, sameness, trust, the polish-versus-avoid line, the agent-list concern.
• Saúl Blanco, Indiana University Luddy School, Advice for Prospective Research Students (no date stated): generative AI, "respected" and "esteemed".
• David Evans, University of Virginia Security Research Group, Advice on Contacting Faculty for Prospective Research Students (no date stated): generic statements, flattery, attachments, follow-up after about a week.
• University of Cambridge Postgraduate Admissions, Using AI in your application (no cycle stated): rules for personal statements and CVs, research proposals and interviews.
• Kobak, González-Márquez, Horvát and Lause, Delving into LLM-assisted writing in biomedical publications through excess vocabulary (Science Advances, July 2, 2025; PubMed abstracts 2010–2024): excess-word ratios for delves, underscores and showcasing; the 13.5% lower bound.