If your application could describe almost any candidate, it needs more of your own evidence. You do not need to prove that anyone else received the same AI wording. The practical problem is that the text gives a recruiter too little to understand your work, your interest in this role or the limits of what you can offer.
Read “sounds like everyone else's” as a useful editing question, not a claim about how many nannies use AI. This guide helps you repair vague generated language. It does not identify AI authorship or offer tricks for passing a detector.
Find the sentences that say very little
Underline phrases that praise you without describing anything: dedicated professional, exceptional care, perfect fit, passion for excellence. Such phrases are not automatically false. They simply leave the reader with the task of imagining what they mean.
Try the substitution test. Could the sentence fit a tutor, assistant or unrelated applicant after changing the job title? If so, ask what information it adds to this particular application. A sentence can be grammatically polished and still fail that test.
Look for generic enthusiasm as well as generic skill claims. “Your prestigious opportunity aligns perfectly with my aspirations” does not explain which aspect of the work interests you. Identify the actual connection before deciding how to phrase it.
Do not solve this by replacing ordinary adjectives with unusual ones. More elaborate wording can preserve the same information gap. Your aim is a clearer account, not a more distinctive disguise.
Diagnose what the draft is missing
For each weak sentence, choose the missing element. Is it missing an action you performed, the context in which you performed it, a relevant preference, or a realistic boundary? Different gaps need different repairs.
A statement about communication might need an example of what you communicated and to whom, described without identities. A statement about flexibility might need the days or arrangements you can genuinely consider. A statement about interest might need a connection to the advertised responsibilities.
OpenAI's prompting guidance recommends clear context and revision after reviewing the response. You can use that approach to request a narrower edit, while supplying only information appropriate to share. OpenAI: Prompt engineering best practices for ChatGPT
If you cannot supply a real example, remove or narrow the sentence. Asking AI to “make it more personal” without adding your own facts does not give it evidence of your experience.
Replace praise with one piece of evidence
Use this original NAJ sentence workshop privately. Write the broad claim on one line. Under it, answer three questions in ordinary language: What did I do? What part was my responsibility? Why is that relevant here?
Then write a sentence from those answers. Avoid including every detail you remember. A useful example may be a routine task performed consistently; it does not need a dramatic story, a named family or an impressive outcome.
Keep the boundary visible. If you supported a plan someone else created, say that. If the work happened occasionally, preserve that frequency. Concrete writing becomes misleading when it removes the context that made the original claim modest and accurate.
For a cover letter, choose the evidence that helps explain this application. The agency cover letter guide covers overall structure. Here, the task is to improve the individual sentences already inside your draft.
Compare a fictional before and after
Original fictional example: Jamie's generated paragraph reads: “I am a highly adaptable childcare professional with exceptional communication skills and a passion for creating enriching experiences. I would be the perfect addition to your household.”
Jamie's own notes say that a previous afternoon role involved following a shared activity calendar and leaving a short handover for the next caregiver. Jamie is interested in another role involving shared care.
A more useful version is: “In my recent afternoon role, I followed the shared activity calendar and left a short handover for the next caregiver about schedule changes and unfinished tasks. The shared-care arrangement in this opening interests me, and I would like to understand how your team organizes handovers.”
The revision makes a specific connection and leaves a genuine question open. It does not claim that Jamie is the perfect candidate or that the same routine will suit every family. The experience is fictional and should not be copied as your own.
Keep a voice you can use in conversation
Read each paragraph aloud. Notice words you would struggle to explain or would never use when discussing your work. Replace them with familiar language that preserves the meaning. Professional writing can sound like a person speaking carefully.
This is not a test of authorship. It is a test of whether you understand and stand behind the text. Someone may use formal language naturally; someone else may prefer shorter sentences. Neither style establishes how a draft was produced.
Ask a trusted reader what they learned about your actual experience. If the answer is only “you sound enthusiastic,” the draft may still need evidence. Ask which sentence felt unclear rather than asking whether it “sounds AI.”
Keep revisions proportionate. You do not need to rebuild an accurate application because one sentence is stiff. Improve the weak section, check it against the rest of the document and stop when the account is clear.
Finish with an evidence and relevance check
Use this original checklist before submitting: each main claim has support; examples describe your contribution; interest connects to the real opening; availability remains accurate; and no household information was added to make the writing vivid.
Check the current application instructions, including any rules about assistance. Save the exact version you submit. If asked how you prepared it, describe the actual assistance truthfully; the AI disclosure guide addresses that separate decision.
The goal is not to make the text look less like a machine wrote it. It is to give the reader something meaningful to understand about you. If a sentence cannot do that, a shorter application may be stronger than another layer of polished language.
Sources
OpenAI: Prompt engineering best practices for ChatGPT, checked October 27, 2026. The sentence workshop, substitution test and fictional example are original NAJ editorial guidance.