Five sentences, one paper, one small ask

What the email is actually for

It is not for getting a research position. It is for getting a fifteen-minute conversation. Every structural decision below follows from that, and most bad cold emails fail because they try to do the whole job in one message: describing a full research interest, requesting a formal role, attaching four documents, and asking for a commitment from someone who has not yet decided you are worth a reply.

Reframe the goal to "earn one short meeting" and the email gets dramatically easier to write. The ask shrinks, the length halves, and the professor's decision changes from "do I want to take on an undergraduate" to "can I spare fifteen minutes for someone who clearly read my paper." The second question is much easier to say yes to.

Before you write anything: pick five labs, not fifty

The single strongest predictor of a reply is whether the email could have been sent to anyone else. Mass mail is detectable in about two seconds, and the tell is not a merge-field mistake. It is the absence of anything specific enough to be wrong.

So the work happens before the writing. Five to eight labs, chosen properly, will outperform fifty generic emails by a wide margin, and the research phase takes an evening.

1. Start with your own department. Proximity matters more than prestige for a first position. A professor two buildings away can meet you Thursday. Someone at another institution cannot, and the logistics alone kill most external arrangements.
2. Read the lab page, then the publication list, sorted by date. You want the last two years. Work from five years ago may not describe what the lab is doing now, and citing an old paper as if it were current is a specific and visible mistake.
3. Read the current students' pages. They tell you what the lab actually works on day to day, what tools it uses, and whether it takes undergraduates at all. A lab with three undergraduate names listed is a different prospect from one with none.
4. Check for a stated policy. Many faculty pages have a line about how to contact them regarding research, sometimes including a specific subject line or a form. Following it is free and skipping it is disqualifying.
5. Rank by overlap, not by fame. The lab whose methods you could contribute to next month beats the famous one whose work you would need two years of background to touch.

Reading the paper properly, in forty minutes

You do not need to understand a paper completely to write a good email about it. You need one specific, accurate observation. The reliable route to one:

• Abstract, then figures, then conclusion. The figures carry most of the actual result and take a fraction of the time of the prose.
• Methods, skimmed for the tooling. What did they build it with, what data did they use, what did they have to do by hand. The hand-done parts are where an undergraduate is genuinely useful.
• Limitations and future work, read closely. This is the most valuable section in the entire paper for your purposes. It is the authors telling you, in writing, what they wish someone would do next. An email that responds to a specific line in future work is nearly impossible to mistake for mass mail.
• One thing you did not understand. A single honest, specific question is better than false fluency, and it gives the meeting somewhere to start.

The test for whether you have read enough: can you name the specific thing the paper did, in one sentence, without using the words "interesting" or "impactful"? If not, read the figures again. If yes, you are ready to write.

The five-sentence structure

Under 150 words in the body. Each sentence has exactly one job, and the order matters because a professor may only read the first two lines before deciding.

Sentence five is the one people skip and it is doing real work. Explicitly giving someone permission to decline raises reply rates, because a busy person who cannot help you now still has a cheap, non-awkward action available. Silence is what happens when the only options are yes or an uncomfortable no.

The subject line

Specific, lowercase-plain, no exclamation marks. What works: Undergraduate interested in your [specific topic] work, or Question about [paper short title], or whatever the lab page explicitly asked for. What does not: Research Opportunity Inquiry, Seeking Research Position, anything that reads like it came from a template, and anything that promises more than the email delivers.

Example one: your own department

The common case, and the highest-probability one. Note the length, the single named paper, the specific skill, and the fifteen-minute ask.

Subject: Undergraduate interested in your bilingual programming instruction work Dr. [Name], I'm a second-year CS major at [University], currently taking [relevant course]. I read your [Year] paper on [exact title] and spent a while on the section where you note that the coding scheme for the open-response items had to be applied by hand across all participants. That constraint is the part I keep thinking about, because it seems to be what caps the study size. I've built a small text-classification pipeline in Python for a course project and I'm comfortable with pandas and basic inter-rater reliability statistics, so the manual coding bottleneck is the kind of work I could actually help with rather than just watch. Would you have fifteen minutes in the next two weeks to tell me what the lab needs right now? My one-page CV is attached. If you're not taking undergraduates this semester, no problem at all — I'd just ask whether there's someone in the group worth talking to instead. Thank you, [Name] [email] · [github.com/username]

What that email does: names a real paper, engages with a specific methodological constraint rather than the topic in general, offers a skill that maps directly onto the constraint, asks for fifteen minutes, and hands over both an easy no and a useful alternative to a no.

Example two: a lab outside your department

Harder, because you have less standing and less overlap. The fix is to be more specific, not more enthusiastic, and to name the logistical reality up front so the professor does not have to raise it.

Subject: Question about the data pipeline in [paper short title] Dr. [Name], I'm a second-year computer science student at [University], not in your department, which I realize makes this an unusual email. I read [exact title] after finding it through [specific, honest route: a citation, a course, a talk]. In the future work section you mention wanting to extend the analysis to [specific thing]. I'd been building something adjacent for a personal project — [one clause on what it does] — and the overlap is close enough that I wanted to ask rather than assume. Concretely, I can write and test Python data pipelines and I've worked with [specific library or method the paper used]. Could I ask for fifteen minutes to hear whether that's useful to you? I can work remotely and I'm not asking for funding. If the answer is no, that's completely fine — I'd still be glad to know if there's a reading you'd point me at. Thank you for your time, [Name] [email] · [portfolio or GitHub link]

Two things that email does deliberately. It names the awkwardness ("not in your department") instead of hoping nobody notices, which reads as honest rather than presumptuous. And it removes the two objections that kill external requests before they are raised: location and money.

What to attach, and what not to

The CV should show the two or three things that make you plausible for this lab, and it should be legible in ten seconds. The same brevity discipline that makes a resume survive a recruiter screen applies here, and it is worked through in the ATS resume guide and the technical portfolio guide. A professor reads it faster than a recruiter does, not slower.

Timing

• Best: the first two or three weeks of a semester, when planning happens, or late in the previous semester for the following one.
• Also good: a few weeks before a summer program cycle, when faculty are thinking about who is available.
• Avoid: finals week, the week grades are due, and the week before a major conference deadline in the field. A well-written email arriving in that window still gets ignored, and you cannot resend it.
• Send Tuesday through Thursday morning. Marginal, but free.

Follow-up: once, then stop

One follow-up, seven to ten days after the original, replying in the same thread so the original is directly underneath. Three sentences maximum, adding something rather than repeating.

Hi Dr. [Name] — following up briefly on the note below in case it arrived at a busy time. Since writing it I [one concrete new thing: finished the related project, read the follow-up paper, took the course]. Still happy to be told this isn't the right semester. [Name]

Then stop. A second follow-up converts almost nothing and costs you the option of writing again next semester, which is a real option worth preserving. No reply is not a rejection; it is usually an inbox. The correct interpretation of silence is "try a different lab now, and this one again in six months with something new to show."

Handling a no

Most nos are logistical rather than personal, and the right response depends on which kind you got. All three deserve a reply, and the reply is short.

The referral ask is the highest-value sentence in this entire guide and it costs nothing: "if this isn't the right group, is there someone in the department you'd point me toward?" Faculty know exactly who is looking for undergraduates and who is not, and a forwarded introduction from a colleague outperforms any cold email you could write.

After a yes: the part that decides whether it lasts

Getting the meeting is the hard part in terms of probability. Keeping the position is the hard part in terms of consequence, because a research role that ends badly is worse than one that never started.

1. Come to the first meeting with two questions and no proposals. You do not yet know enough to propose a direction, and trying to signals the opposite of what you intend.
2. Ask for a small first task and finish it early. The first deliverable is a reliability test, not a talent test. Finishing a small thing on time establishes more than doing a large thing brilliantly and late.
3. Write down what you did, weekly. A short log becomes your contribution record, your progress-report material, and eventually the evidence base for a recommendation letter. Nobody regrets keeping one.
4. Say when you are stuck, within a day. The failure mode that ends undergraduate research positions is not being stuck. It is being silently stuck for three weeks.
5. Ask about authorship norms early and directly, once. It is a normal question, and asking about it in month one is far better than assuming for a year.

That trajectory is what turns a cold email into something that appears on a resume, and eventually into published work: my own path into a peer-reviewed SIGCSE Technical Symposium 2026 paper with the SJSU CSEd Research Lab under Dr. Ethel Tshukudu started as exactly this kind of conversation, and the first task was small and unglamorous. What the position actually looks like from the inside, and how to choose between labs once you have options, is covered in the undergraduate research guide and on the research page.

The killers, listed plainly

• Wrong name, wrong lab, wrong paper. Instant delete, and correctly so. Proofread the salutation last, every single time.
• "I am passionate about your research." Every one of the forty emails says this. None of them are distinguishable because of it.
• Obvious mass mail. No specific paper, no specific method, nothing that could not have been sent to any professor in the department.
• Length. Over 200 words in the body and the reply rate drops sharply. There is no version of this email that needs three paragraphs of background.
• Asking for too much. A recommendation letter, a paid position, or co-authorship in a first email are all requests from a stranger, and they read that way.
• Attachments nobody asked for. A five-megabyte folder of certificates is a burden, not a credential.
• Generated text left generic. Drafting with a tool is fine. Sending something that reads as unedited template is the same failure as mass mail, with extra steps.
• Not following the stated instructions on the lab page. If the page asks for a specific subject line and you did not use it, you have already answered the question of whether you read carefully.

The realistic numbers

Six carefully researched emails commonly produce one to three replies and often one meeting. Forty generic ones commonly produce zero. That is not a motivational claim, it is a consequence of the filter: the only thing being evaluated in the first two seconds is whether this email could have been sent to someone else.

And if a first round returns nothing, the response is not a longer email. It is more specific reading, a smaller ask, and a better-chosen list, then a second round next term with one new thing to show. The same discipline of tracking who you contacted, when, and what came back applies here exactly as it does to job applications, which is what the Apply OS tracker is for.

Tools referenced in this guide

• Research — the SIGCSE 2026 paper and the lab work behind them.
• Undergraduate research guide — what the position looks like from the inside, and how to choose between labs.
• Apply OS — tracking who you contacted, when, and what came back.
• Resume — the one-page format that attaches to this email.