NSF GRFP Examples: Winning Personal Statements and Research Proposals

NSF doesn't publish past applications. Every example you can find online was shared by the person who wrote it, either on their own site or through a collection or a university writing center. Here are the collections worth your time, listed roughly in the order you should read them.

The next four sections list all 99 applicants, with 220 documents from 2014 to 2026, sorted by NSF's major fields. In the tables, "Statement" means the personal statement and "Plan" means the research proposal. "Reviews" means the writer also shared NSF's ratings sheet, with each reviewer's comments. Most of these applicants won the fellowship, and 8 are clearly labelled honorable mentions.

Pay attention to when each person applied. Second-year graduate students can't apply anymore. Their statements also describe more experience than a senior or first-year can usually show, so read those for structure only.

Read a dozen of these and you'll notice the same habits, whatever the field. On GRFP Help, all 36 research plans and all 36 statements with searchable text use separate Intellectual Merit and Broader Impacts headings. Of the 36 plans, 21 use the word hypothesis, and 16 label numbered aims, objectives or goals. The excerpts below show the habits that matter most. Borrow the move, but never the words.

Of the applicants above, 47 shared their NSF ratings sheets. Read a sheet next to its essays to see what the panel noticed. Even winning applications got criticism. And an honorable mention can earn Excellent ratings and still miss out on funding.

NSF describes honorable mentions as "meritorious applicants who do not receive Fellowship offers" and calls the recognition "a significant national academic achievement" (NSF 26-526). But it doesn't come with any money. For the 2026 competition, NSF made 2,500 offers and named 1,470 honorable mentions (NSF FAQ).

The two essays do different jobs, but reviewers read them together. Your personal statement makes the case that you'll become a scientist who advances knowledge and benefits society. Your research plan shows how you think through one original question. Here's what the current solicitation and NSF's own templates ask for.

Reference letters are due October 16, 2026. Applications are due October 19 to 23, depending on your field, at 8:00 p.m. Eastern (NSF 26-526). If you're a senior, or a graduate who has never enrolled in graduate school, you can apply any number of times. First-year graduate students can apply once. The deadlines guide lists every date, and the eligibility guide explains who can apply. For anything else, NSF's GRFP program page has FAQs and contacts.

Most public essays were written under older rules. You can still learn a lot from how they argue, but their formatting and eligibility may be out of date. Check each example against these changes before you borrow a habit from it.

• Headings. MIT's Communication Lab notes that applications before 2018 "did not require" separate Intellectual Merit and Broader Impacts headings (MIT CommLab). Now they're required, and NSF adds: "Do not include other terms. Do not combine the headings or the sections."
• Format before 2014. Alex Hunter Lang's collection marks 2013 and earlier as the old format, which had a third essay on previous research. Read those for voice, not layout.
• Second-year graduate students. Many older examples come from second-years, who haven't been eligible since the 2025–26 cycle. Their statements include more research than a first-year can usually show.
• Fields and topics. STEM Education and Learning Research is gone as a field. Clinical and counseling psychology are out. Conservation practice research is now excluded too.
• Deadlines. The dates move every year. Last cycle's deadlines were November 10 to 14 at 5 p.m. local time. This cycle's are October 19 to 23 at 8 p.m. Eastern.
• References and figures. NSF now requires "abbreviated journal or publication titles ONLY, no URLs or DOIs", plus alt text for every figure. Brandon Sike's 2025 plan, for example, lists DOI links that wouldn't pass today.

NSF makes you certify that both statements are "their own original work" (NSF 26-526). Reading winning essays is normal and really useful. Reusing their sentences is not.

1. 1

   Pick three recent ones from your field

   Panels are organized by field, and each field has its own norms. A chemistry plan leans on mechanisms and figures. An economics plan leans on data and identification.

2. 2

   Add one from another field

   A strong plan from outside your area shows you what reads clearly to a non-specialist. That's who most of your panel will be.

3. 3

   Map the structure

   Note the headings, how much space each part gets, and where the research question first shows up. MIT's annotated research plans walk through this part by part.

4. 4

   Mark the evidence

   Underline every number, named program, result and "I". Aim for that much evidence in yours, using your own facts.

5. 5

   Read the reviews next to the essays

   If a ratings sheet is posted, match each comment to the part of the essay that earned it.

6. 6

   Close the examples, then write

   Draft from your CV and your own notes. The personal statement writer and research proposal writer build drafts from your material, never from sample essays.

• Copying the opening. A borrowed hook sounds borrowed. Start with a moment of your own, or with your research question.
• Copying old formatting. Some past winners use combined headings, headings that run straight into the paragraph, or DOI links in their references. None of that passes now. Check your PDF with the format checker.
• Matching length instead of substance. Winners fill their pages with evidence, not adjectives. Cut any sentence a reviewer couldn't check.
• Reading only your subfield. Your panel is broader than your lab. Write so that a scientist down the hall could follow you.
• Planning outreach you've never done. Strong Broader Impacts build on something you've already started. They name a partner, a place and a number.
• Treating one example as the formula. Some winners open with a story, some with a goal and some with a question. What they share is evidence, not a template.