Words to Avoid in an SOP: Clichés and AI Tells

Words to Avoid in an SOP: Clichés and AI Tells

Do not ban words; replace claims. Any adjective or adverb in your statement of purpose that a reader cannot check against your CV, transcript or the program's own page (passionate, world-class, crucial, comprehensive) should become the fact behind it, or go.

Our view: the cliché and the AI tell have one cause, a claim without evidence. That is why a list of banned words fails. Swap passionate for deeply enthusiastic and the sentence is still empty. Replace it with what you did for two years, and there is nothing left for a reader to doubt. This page gives the three families of empty phrases, the AI vocabulary that a large study actually measured, and the replacement rule.

Every university page quoted below was checked on October 31, 2026.

Three families of empty phrases

Each row is a constructed example we wrote, not a quote from an applicant.

Claims about yourself

Praise of the program

Indian English has its own flattery phrases ("esteemed university", "reputed institute"); Indian SOP clichés covers them, along with the "since childhood" opening.

AI style words

The AI vocabulary, measured

Here is what one large study actually measured. Kobak and colleagues analyzed "over 15 million biomedical abstracts from 2010–2024 indexed by PubMed" and compared each word's 2024 frequency with the frequency expected from earlier years (Science Advances, July 2, 2025; arXiv version).

What they found, in their words: "Less common words with strong excess usage included delves (r = 28.0), underscores (r = 13.8), and showcasing (r = 10.7)". Here r is the ratio of actual to expected frequency, so delves appeared 28 times as often as expected. "More common words with strong excess usage included potential (δ = 0.052), findings (δ = 0.041), and crucial (δ = 0.037)". Here δ is the gap between the actual and expected share of abstracts containing the word, so potential appeared in about 5 percentage points more abstracts than expected (our arithmetic: 0.052 = 5.2 points).

They also built a set of 10 common style words: "across, additionally, comprehensive, crucial, enhancing, exhibited, insights, notably, particularly, within." And they found that "out of all 379 excess style words in 2024, 66% were verbs and 14% were adjectives." Their overall estimate: "at least 13.5% of 2024 abstracts were processed with LLMs."

What the study does not show:

• It is about biomedical abstracts, not statements of purpose. No SOPs were in the data.
• It cannot judge one text. The authors say: "Our analysis is performed on the corpus level and cannot identify individual abstracts that may have been processed by an LLM." One delve in your draft proves nothing.
• Some of these words are field terms. Potential in physics and findings in a research summary are ordinary vocabulary. Keep them where they name a real thing.

Use the list as a prompt to look again at a sentence, not as a ban. Your SOP sounds AI-written explains why a statement you wrote yourself can still read that way.

What university readers say about empty phrases

Three readers' pages say it directly. The University of Toronto MScAC program tells applicants that AI tools "usually do not do a good job emphasizing your unique strengths, and do not differentiate you from other candidates who also used such tools." Cambridge's postgraduate guidance asks applicants to "include relevant examples of your experience, do not just cram everything in" and to "be honest about your achievements and experience."

The University of Virginia security research group's page for prospective students, written about emails asking to join the group, puts the flattery problem bluntly: "Don't waste space and time telling me how brilliant I am." And on self-praise: "Don't waste space and time telling me how hard-working, creative and smart you are — demonstrate it with the contents of your message." That page is about emails, not SOPs; we quote it because it is a faculty reader's own words on flattery and self-praise.

The replacement rule

Run this on every paragraph:

1. Underline every adjective and adverb about you, your work or the program.
2. Ask of each one: could a reader check this against my CV, transcript or the program's page? Pass: yes, and the evidence is in the same sentence.
3. If not, replace it with the fact behind it: a number, a name, a date, a decision. If there is no fact behind it, delete it.
4. For AI style words, replace the verb with what actually happened: not "delved into" but "tested", "measured", "rebuilt".
5. Read the paragraph aloud. If a sentence could appear in another applicant's statement unchanged, it still needs a fact.

Constructed example: one paragraph, six empty phrases

Constructed example, written by us, not a real applicant statement.

Before (invented, 86 words, not counting the numbered markers):

I am a passionate and hardworking student (1) with a keen interest in data science. Your world-class faculty and cutting-edge research (2) make this program the right place for me. During my internship at a retail bank, I delved into a crucial project (3) that showcased my analytical skills (4). Notably, I gained comprehensive exposure to real-world data (5) and worked with many teams. I am confident that I will be a valuable addition to your program (6), and that it will help me reach my goal of becoming a data scientist.

After (invented, 95 words):

During a six-month internship at a retail bank, I rebuilt the monthly customer-churn report that three regional teams used to decide whom to call. The old report used last year's averages. I replaced them with a logistic model on 40,000 accounts and gave the teams the 2,000 customers most likely to leave. Two teams adopted it. I could not explain to them why the model ranked some customers high, and that gap is why I am applying: the program's Interpretable Machine Learning course and its industry capstone are the training I do not yet have.

What changed: (1) became the internship and its length; (2) became one named course and one requirement; (3) and (4) became what the applicant built and who used it; (5) became the dataset size; (6) became a gap the program fills. The rewrite is nine words longer and carries every fact the first version only claimed.

Do not swap one list for another

Synonyms keep the claim empty. "Passionate" becomes "deeply enthusiastic"; "delve" becomes "dive deep". Nothing changes for the reader. Paraphrasing or "humanizer" tools do the same thing at scale, and on a statement a tool wrote, they leave the tool's argument in place. Do AI humanizers work? covers the evidence; rewrite a ChatGPT SOP in your own words covers the rebuild.

If you are tempted to let a tool polish your wording, read Grammarly and ChatGPT rules for SOP English first.

The countercase: some words on these lists are the right word. A research summary has findings; a chemistry statement may need potential. A writer whose sentences already carry a fact beside every adjective does not need to strip them. The rule targets claims without evidence, not vocabulary.

Questions about SOP word choice

Is "passionate" banned in a statement of purpose?

No university page we checked bans a word. The problem is that "passionate" is a claim the reader cannot check. Replace it with what you did; if the evidence is strong, you will not need the adjective.

Can I use "delve" if I wrote it myself?

Yes. The Kobak study works "on the corpus level and cannot identify individual abstracts". One word is not evidence of anything. Use it if it is the exact word you mean. If a plainer verb says the same thing, use that.

Should I use advanced vocabulary to impress the committee?

No. The pages quoted here ask for examples and honest evidence, not rare words. Simple English SOP for non-native writers has a longer cut list and seven rules for plain, exact sentences.

Will removing these words stop my SOP from being flagged as AI?

We cannot tell you that, and no source we found says so. Replacing claims with facts makes the statement stronger whether or not anyone runs a tool on it. The six deadliest SOP mistakes covers the other problems that make a statement read as generic.

Get a free review of your statement of purpose

Get a free Quick Review of your statement of purpose against GradPilot's general master's SOP rubric. It reads for specific academic goals, evidence of preparation and fit with the program. The scholar voice rubric goes further on wording: it checks whether the sentences about your own work could belong to another applicant unchanged. You can also start from the master's SOP review page and choose either rubric there.

More guides on SOP wording:

• Indian SOP clichés: "esteemed university", "since childhood" and the formal-letter register.
• Why your SOP sounds generic: the sentence-level cause.
• All graduate school essay guides.

Sources

Retrieved October 31, 2026, except where noted.

• Kobak, González-Márquez, Horvát and Lause, Delving into LLM-assisted writing in biomedical publications through excess vocabulary (Science Advances, Vol. 11, No. 27, July 2, 2025; PubMed abstracts 2010–2024; full text from the arXiv PDF, version 5): excess-word ratios and gaps, the 10-word set, the 13.5% lower bound, corpus-level limitation.
• University of Toronto MScAC, Apply (no year stated): guidance on AI tools in essay answers.
• University of Cambridge, Writing a personal statement (no year stated): tips for writing a statement.
• University of Virginia Security Research Group, Prospective students (no year stated): guidance on emails from prospective students.