Thank you. 500+ students, 600+ essays this season. 25% off with NUDGE25 through Oct 5.

How to write a college essay, from seven marked-up reports

To write a college essay: choose a subject whose details could not be transplanted into anyone else’s draft, spend most of the words inside the moment rather than on the meaning of it, and stop on a specific image instead of a summary of what you learned. Those three moves are what separate the strongest and weakest essays in the only evidence this page uses, which is seven complete reports the NudgeEssay engine wrote on seven sample essays and published in full and unedited.

The pattern in those seven: the two essays whose sentences the engine found least replaceable sit inside no stock archetype at all, and the two least distinguishable both sit deep inside a named one. The four personal statements rated competent or better put 45 to 48 percent of their words in scene; the two rated lowest put 0 and 5 percent there. In four of the seven reports the first-ranked fix is a cut in the closing paragraphs. Seven essays is not a study, this page never claims more than the seven show, and nothing here predicts what any college will decide.

Updated . Written and maintained by NudgeEssay; how facts on this site are sourced and corrected is on the about page.

Choosing a topic: the shapes a reader has already seen

Every NudgeEssay report names the nearest known stock shape the draft is sitting inside and how deep in that groove it sits. It also gives an Only-You index: the report’s own estimate, out of 100, of what share of the essay’s sentences no other applicant could have written. Across the seven published reports, the two readings line up like this.

One-week test, detected archetype and Only-You index for each of the seven published NudgeEssay sample reports

The freezer in the garage

One-week test
exceptional
read the report
Archetype detected
none
Groove risk
none
Only-You index
88

The loft

One-week test
memorable
read the report
Archetype detected
none
Groove risk
none
Only-You index
82

The penalty kick

One-week test
memorable
read the report
Archetype detected
the big game won or lost
Groove risk
medium
Only-You index
52

The co-op ledger

One-week test
competent
read the report
Archetype detected
none
Groove risk
none
Only-You index
62

Three lunchrooms

One-week test
competent
read the report
Archetype detected
moving schools / new-kid
Groove risk
medium
Only-You index
58

The tapestry

One-week test
forgettable
read the report
Archetype detected
immigrant-grandparent tribute
Groove risk
high
Only-You index
3

Her last gift

One-week test
working against the applicant
read the report
Archetype detected
a death used as a trophy
Groove risk
high
Only-You index
8

Every figure is read at build time from the published fixture behind each report, which is unedited engine output. Seven reports is the whole gallery and the whole evidence base for this page. They were picked to span the range, so they are not a sample of applicants’ drafts and no rate should be read off them.

three of the seven map to no stock shape at all, and they include both of the essays whose sentences the engine found least replaceable, at 88 and 82 out of 100. Of the pigeon loft essay the report says “this essay maps to no known dead archetype; it is genuinely its own thing”. Of the roadkill essay, the only exceptional in the gallery, “roadkill articulation with a state salvage permit and a beetle colony is genuinely sui generis, and the reflection does not reach for a borrowed lesson”.

The two at the bottom, at 3 and 8, both sit deep inside a named one. On the grandmother essay, filed under a death used as a trophy: “the grandmother's death functions primarily as the engine of the student's achievement list rather than as a relationship explored on its own terms; the only escape is to make the grandmother a specific, living person in the reader's mind before she is gone”. On the tapestry essay: “it hits every expected beat (two worlds, translator metaphor, resilience, growth) without a single detail that makes this student's version of the story different from the ten thousand others; the only survivable path is a specific, unexpected scene that no one else could have written”.

The honest reading is not a list of banned topics. The penalty-kick essay sits squarely in the big game won or lost at medium risk and still came back memorable, because the report found the real subject is elsewhere: “The essay survives the archetype because the penalty kick is not the real subject”, it says, “the taping ritual and the absent father are”. The cost of the groove is only that “the opening paragraphs will read as familiar until the taping detail arrives”.

And a clean archetype read is not enough on its own. The Why Us essay is filed as “it is a genuine Why Us built from a real work history, not a narrative archetype” and still landed at competent, for an unrelated reason: “the absence of a single named Northeastern-specific fact (a co-op site, a professor, a course) means nothing sticks after the pile”. Escaping the stock shape is necessary, not sufficient.

So the question to ask about a topic is not whether it is common. It is the one the grandmother report ends on: “Read the revised draft and ask: 'Could a different student who also lost a grandparent have written every sentence of this?' For every sentence where the answer is yes, replace it with something only you could have written.”

When the answer is that anyone could have, both weak reports prescribe the same exercise, and it is the most concrete topic-finding instruction in the gallery. One asks for 100 words, tonight, no editing and no polish, “describing one specific moment from the day of the fundraising walk: where you were standing, who was there, what you were holding, what you were thinking. Do not use the words 'resilience,' 'strength,' 'purpose,' or 'memory.'” The other asks for fifteen minutes: “Tonight, write for 15 minutes answering only this: describe the single most specific moment when being between two cultures required you to make a choice” … “what did you see, hear, say, or do? Do not use any metaphors. Just say what happened.” The forbidden-word list is the mechanism in both. Every word on it names a category a reader has already filed.

What to cut, and what a screen actually flags

Each report carries an AI-pattern screen, and it is a screen readout rather than an accusation: it reports how the prose would look to the machine screens some colleges now run, not what anyone did. Across the seven reports, three came back low, two medium and two high.

The two medium readings are the useful ones, because both name the same block of prose. On the three-lunchrooms essay: “The final paragraph ('Moving so often taught me things I would not trade... resilience and adaptability will serve me well in college') reads with the smooth, generic cadence an AI-pattern screen would flag”, and then, plainly, “The rest of the essay is low-risk and distinctly voiced.” On the penalty-kick essay it is the final two paragraphs that “are smooth in a way that outpaces the rest of the essay's register and would flag on an AI-pattern screen; the earlier paragraphs (the taping ritual, the quarterfinal tackle) read as authentically teenage and protect the overall score”. In both cases the story is clean and the summing-up is what would trip the screen.

The two high readings belong to the two essays with almost no story in them at all, and there the screen names the whole draft. On the tapestry essay: “unnaturally even transitions ('This navigation between worlds instilled in me'), abstract emotional language with zero sensory grounding, and a closing paragraph structure ('My mosaic is far from complete… only enriched') that is a textbook AI essay conclusion”. On the grandmother essay: “The transitions ('But then I remembered,' 'From that moment on,' 'As I prepare for college'), the three-item lesson lists ('resilience, dedication, and the true meaning of perseverance'), and the closing paragraph's cadence ('every classroom, every exam, and every challenge') are all high-frequency AI-pattern constructions”. What protects an essay, by contrast, is unglamorous fact. The loft report lists “the numbered birds, the rest-stop parking lot at 5 a.m., the spiral notebook labeled LOFT in shaking capitals” and concludes: “no AI generates that density of specific, unglamorous fact. A machine screen would flag nothing.”

The generic-sentence lists point at the same place. Five of the seven reports say in their own notes that the generic sentences sit in the reflection or the final paragraph: “The two generic sentences are both in the final paragraph”, “the three generic quotes are all in the reflection paragraphs, which is the expected pattern”, “the reflection paragraphs pull the index down by retreating to language any athlete could have written”. Even the essay with the highest Only-You index in the gallery has generic sentences listed, and both are in its last paragraph. The lines themselves all do one thing, which is name a virtue instead of showing one: “belonging is something you build, not something you find”, “It taught me resilience, dedication, and the true meaning of perseverance.”, “Curiosity became my compass, guiding me through the vast landscape of knowledge.”

What the reports mark as strength is always a physical fact carrying its own weight. On the loft essay the marked sentence is “four ounces underweight with a healed tear in one wing”, and the note reads “This is the sentence a reader carries into committee. 'Four ounces underweight with a healed tear in one wing' is the kind of physical fact that makes a story real.” On the penalty-kick essay it is “Two wraps around the arch, figure eight, two anchors.”, marked best image in the essay because “the numbered steps read as ritual, the four-minute precision is oddly moving, and 'completely calm' lands because it is understated”. On the three-lunchrooms essay it is “doing the math on every table in the room”, because “'doing the math' captures the social calculus of adolescence better than any abstract description could”. And on the Why Us essay it is “at 6 p.m., when the line is long”: “The 'again' and the '6 p.m.' are the two words that make this sentence real. This is the only moment in the essay where the reader is inside the pharmacy, and it earns everything that follows.”

There is a matching warning about cutting too much. The one report that came back exceptional opens by saying that “the single biggest risk is not fixing it”, “it is over-fixing it”. Even its one flagged comma splice is hedged, “though the run-on rhythm may be intentional”. A revision that sands off every rough edge also removes the reason the essay was worth reading.

Structure: where the words actually go

Every report classifies each sentence as setup, scene or reflection and prints the three percentages. That split, far more than length, is what these reports keep returning to.

Setup, scene and reflection percentages for each of the seven published sample reports

The freezer in the garage

Essay type
Personal statement
Setup
22%
Scene
45%
Reflection
33%
One-week test
exceptional
read the report

The loft

Essay type
Personal statement
Setup
18%
Scene
47%
Reflection
35%
One-week test
memorable
read the report

The penalty kick

Essay type
Personal statement
Setup
22%
Scene
48%
Reflection
30%
One-week test
memorable
read the report

The co-op ledger

Essay type
Why Us
Setup
18%
Scene
22%
Reflection
60%
One-week test
competent
read the report

Three lunchrooms

Essay type
Personal statement
Setup
18%
Scene
47%
Reflection
35%
One-week test
competent
read the report

The tapestry

Essay type
Personal statement
Setup
55%
Scene
0%
Reflection
45%
One-week test
forgettable
read the report

Her last gift

Essay type
Personal statement
Setup
30%
Scene
5%
Reflection
65%
One-week test
working against the applicant
read the report

Percentages are the engine’s own classification of every sentence in each essay, taken from the published report. They are whole numbers and do not always sum to exactly 100.

The four personal statements rated competent or better cluster tightly: 45 to 48 percent scene, with setup held to 18 to 22 percent. The loft report calls that split the right one for the form and says why: “the reflection earns its 35% because it arrives after concrete evidence, not before it; the setup is lean enough that the essay never stalls”.

The two weakest invert it. The tapestry essay scores zero percent scene, and the report treats that as the entire diagnosis: “Zero percent scene is the diagnosis: this essay is entirely setup and reflection with no story in between, which means the reflection has nothing to reflect on and the setup has nowhere to go.” The grandmother essay has five percent scene against sixty-five percent reflection: “this draft tells a reader what to feel rather than creating the conditions for them to feel it; the revision goal is to flip scene and reflection closer to 40/40, using the freed setup words to build one real moment”.

The right split depends on the essay type, and the reports say so. The Why Us essay is sixty percent reflection and the report calls that correct: “The essay is reflection-heavy by design, which is appropriate for a Why Us essay”, with the caveat that “the one scene (the amoxicillin/dosing syringe moment) is doing a lot of work alone”. A short school-specific answer is an argument. A personal statement is a story.

Openings. five of the seven were marked strong and two throat-clearing. Every strong one opens on a concrete, slightly odd fact: “My grandfather numbers his pigeons instead of naming them, which I used to think was cold.” (“Numbering instead of naming is a genuinely strange opening fact, and "the part nobody films" lands before the first paragraph is over”), “I pick up dead animals off Route 9. That sounds like a confession but it's a hobby, or maybe a discipline.” (“Opening sentence is a claim that creates immediate tension and specificity”), “I can tell you the exact layout of three different high school lunchrooms, which is a strange expertise, but it is mine.”, and “Two hundred people watched me place the ball on the penalty spot in the state semifinal, and none of them knew that I almost was not on the field at all.”. Both flagged openings are a general truth about life: “They say that everything happens for a reason.”, of which the report says “it signals immediately that the essay will follow a familiar arc, and it does”, and “Identity is not a monolith but a mosaic, and mine has been assembled from countless fragments of experience”, where “Opening with an abstract aphorism about identity being a mosaic signals immediately that the essay will traffic in metaphor rather than scene, and the skim begins at word one.”

Length. The Common App personal statement caps at 650 words, and the seven sample essays run from 248 to 547. Supplements are far shorter, and these figures come from NudgeEssay’s own prompt dataset rather than from any guide: across the 58colleges checked against each college’s own admissions site whose every prompt states a limit in words, the median college asks 500 words of supplemental writing in total, the most common single cap is 250 words, and the most any one college asks is 1,650 words at Texas A&M University. The full breakdown, college by college, is on the supplemental-essay data page, and your own colleges’ prompts and limits are on the requirements pages.

Making it memorable: the one line a reader keeps

Every NudgeEssay report opens on the memory line: the single sentence a reader could still say a week later, with no notes, after four hundred more files. It is the clearest test on this page, because the strong and weak versions can be set side by side and the difference is visible in one read.

“bird 2119 walked back across two states nineteen days after being logged as lost, and they're inheriting the loft and the shaking-hands grandfather's spiral notebook this spring”

The loft, memorable

“The tape job their father taught them at ten … carried them through a sprained ankle to a state semifinal penalty kick that went in off the keeper's glove”

The penalty kick, memorable

“Three lunchrooms, three states, one method: they cracked the new-kid problem by looking for a job instead of an empty seat, and the one-hit-wonders trivia save in New Jersey is the detail I'd repeat out loud.”

Three lunchrooms, competent

“Grandmother died of cancer sophomore year, and they went from 3.5 to 3.9 GPA, volleyball captain, NHS president, and a $4,000 cancer walk … the résumé is real, but the essay could be anyone's”

Her last gift, working against the applicant

“A bicultural applicant whose essay about identity as a mosaic left no single scene, name, object, or moment I could carry out of the room.”

The tapestry, forgettable

The three that survive are made of physical facts: a weight, a number of days, a taping sequence, a trivia question about one-hit wonders. The two that do not are made of categories. The grandmother essay is the sharper lesson of the pair, because its facts are objectively impressive, a grade point average, a captaincy and a four-thousand-dollar fundraiser, and the reader still finished with “the résumé is real, but the essay could be anyone's”. On the tapestry essay the memorability band reads: “A tired reader who finishes this essay has no image, no name, no moment, no line to carry into committee.”

Memory is most often lost in the last paragraph. In four of the seven reports the first-ranked fix is a cut in the closing, and a fifth pairs a cut of the final paragraph with building a scene. The three-lunchrooms report is the plainest: “Cut or gut the final paragraph (all six sentences): this is the single highest-payoff edit in the essay”, because “it removes the only generic language, raises Authentic Voice, Reflection, and Structure bands simultaneously, and lets the dining-hall closing image land as the emotional and thematic conclusion it was always meant to be”. Its memorability band gives the cost of leaving it in: “What prevents a 5 is that the final paragraph actively dilutes the memory by replacing those images with abstraction.”

The practical version, and it is the same instruction in four of these reports: find the last real image in your essay, then delete everything after it that explains what the image meant.

What seven essays cannot tell you

This page argues from sevenreports and one prompt dataset, and it should not be read as more than that. The seven were chosen to span the range, from exceptional to working against the applicant, so they are not a random sample of applicants’ drafts and no percentage or rate should be read off them. They are all output from one engine, so where the reports agree with each other that is a consistency, not an independent confirmation. No accuracy figure appears anywhere on this page because none has been measured.

Nothing here says anything about admission. NudgeEssay is not affiliated with any college, cannot predict or influence a decision, and does not write or rewrite essays. What this page can offer that a general guide cannot is the reports themselves: all seven are published in full and unedited, free to read with no account, so every claim above can be checked against the sentence it came from. The inventory of every check a report runs is on what is included, and what software can and cannot read is set out on the AI review page.

If you have not chosen what to write about yet, that is a different job from anything above and it has its own page: the free essay topic quiz.

How to write a college essay, answered

How do I write a college essay?
Pick something whose details could not be transplanted into another applicant's draft, write the moment rather than the meaning of it, and stop on a specific image instead of a summary of what you learned. Across the seven complete reports NudgeEssay publishes, the four personal statements rated competent or better put 45 to 48 percent of their words in scene, and the two rated lowest put 0 and 5 percent there. In four of those seven reports the first-ranked fix is a cut in the closing paragraphs. Seven essays is not a study, and no page can tell you what any college will decide.
What should I write my college essay about?
Something a reader cannot file under a shape they have already seen a thousand times. Every NudgeEssay report names the nearest stock archetype and how deep in the groove the draft sits. In the seven published reports, the two essays with the highest Only-You index (88 and 82 out of 100) sit inside no archetype at all, and the two lowest (3 and 8) both sit deep inside a named one at high risk: a death used as a trophy, and an identity tribute. The topic itself is not the disqualifier. One sample about a state-semifinal penalty kick sits squarely in "the big game won or lost" and still came back memorable, because the real subject turned out to be an ankle-taping ritual the writer's father taught them.
What should you avoid in a college essay?
The summing-up at the end. In the two sample reports where the AI-pattern screen came back medium rather than high, the flag is confined to the final paragraphs and the reports say so plainly: "The rest of the essay is low-risk and distinctly voiced." Five of the seven reports also locate the generic sentences in the reflection or the last paragraph, and those sentences all do the same thing, which is name a virtue instead of showing one: "belonging is something you build, not something you find", "It taught me resilience, dedication, and the true meaning of perseverance", "the discipline to prepare when nobody is watching". Avoid opening on an aphorism too. Both sample openings marked throat-clearing began with one.
How long should a college essay be?
The Common App personal statement caps at 650 words, and the seven sample essays here run from 248 to 547. Supplements are much shorter. Across the 58 colleges NudgeEssay checked against each college's own admissions site whose every prompt states a limit in words, the median college asks 500 words of supplemental writing in total, the most common single cap is 250 words, and the most any one college asks is 1,650 words at Texas A&M University. Length is rarely the problem these reports find. The split between scene and reflection is.
How do I start a college essay?
With a fact only you could state. five of the seven sample openings were marked strong, and every one of them opens on something concrete and slightly odd: numbered pigeons, dead animals on a named road, the layout of three lunchrooms, two hundred people and a withheld secret. The two marked throat-clearing both open on a general truth about life, and one report says of its opening that "the skim begins at word one". If your first sentence would survive being pasted into someone else's essay, it is not your first sentence.
What makes a college essay memorable?
One sentence a reader can still say a week later with no notes. Every NudgeEssay report opens with that line. On the strongest sample it is a bird that "walked back across two states nineteen days after being logged as lost". On the weakest it is a grade point average, a captaincy and a dollar figure, followed by the reader's own verdict: "the résumé is real, but the essay could be anyone's". The difference is not how impressive the facts are. It is whether the sentence is made of physical detail or of a category the reader has filed before.
Can I get my own college essay read this way?
Yes. NudgeEssay runs the same engine that produced the seven published reports on your own draft and returns the same report: the memory line, the archetype read, the AI-pattern screen, the setup, scene and reflection split, six rubric bands, notes on your exact sentences, and a fix list ranked by payoff. It is $49 for one essay or $299 for a pack of 20, and you see the verdict on your own essay free before deciding whether to pay. It never writes or rewrites the essay, and it makes no claim about any admissions decision.