You are a hiring manager for {{target_role}}. Using the job description and my existing bullet points, rewrite each bullet to:
- Keep the same task and scope (do not invent anything)
- Add a metric only where I already gave a number, or leave a [blank] for me to fill
- Work in one or two keywords from the job description per bullet
- Keep each bullet under two lines
Job description:
{{job_description}}
My experience:
{{experience}}
2. 讓每一條條列都值得存在
弱的條列描述職責，強的條列描述成果。最可靠的公式來自 Google 前人資長 Laszlo Bock：「以 [Y] 衡量，透過做 [Z]，完成了 [X]。」講白話，就是一個有力的動詞、一個數字，加上你怎麼做到的。
Rewrite each of these into a resume bullet using the formula "Accomplished [X] as measured by [Y] by doing [Z]." Start each one with a different strong past-tense verb. Use only the facts and numbers I give you. Never invent a metric. If a number is missing, leave a [blank].
My notes:
{{rough_notes}}
You are an experienced recruiter in {{industry}}. Write a 3-line professional summary for me that:
- Matches the target role: {{target_role}}
- Includes two or three hard skills from the job description
- Mentions one or two results using only numbers already in my resume
- Uses plain language, no buzzwords
Return only the summary.
Job description: {{job_description}}
My resume: {{resume}}
From this job description, list the hard skills, tools, and certifications it asks for. Compare them to my resume and return a table: Skill | In the job ad? | On my resume? | Action (add / keep / ignore). Only suggest adding skills I genuinely have.
Job description: {{job_description}}
My resume: {{resume}}
5. 要一份誠實的評論
在送出之前，讓 ChatGPT 扮演那個挑剔的人：
Act as a recruiter hiring for {{target_role}}. Compare my resume to this job description and score the match from 0 to 100. Then list: (a) the five most important requirements I'm missing, (b) bullets that read as vague or generic, (c) anything that sounds AI-written. Do not rewrite anything yet. Just give me the audit.
Job description: {{job_description}}
My resume: {{resume}}
這是用 AI 寫履歷最危險的部分，而且它跟被軟體「抓到」一點關係都沒有。陷阱在面試。招募人員讀到「20% 營收成長」，產生興趣，然後請你帶他走一遍。如果你說不出來，你就剛剛當面對那個唯一重要的真人證明了：你在履歷上灌水。
解法就是你加進每個提示詞裡的一句話：
Use only the experience, skills, and numbers I provide. Never invent a metric, tool, or accomplishment. If a number is missing, leave a [blank] for me to fill.
就算事實都是真的，未經處理的 AI 產出還是有股味道。招募人員到現在已經讀過幾千份這種東西了，他們很快就會比對出模式。幾位具名的招募人員告訴 HuffPost，到底是什麼讓他們起疑：同樣的那幾個字、同樣的節奏、同樣的範本句。
那些洩底的字眼
ChatGPT 有幾個偏愛的動詞。「Spearheaded」是最大的一個（有個履歷團隊稱它是「每個主流 LLM 的絕對最愛動詞」）。「Leveraged」、「orchestrated」、「seamlessly」和「at the intersection of」這個片語也是。像「results-driven」、「dynamic」、「passionate」、「cutting-edge」這種流行詞，效果一樣糟。當你發現它們時，換成你真正想講的那個樸實動詞：不是「leveraged Salesforce」而是「used Salesforce」；不是「spearheaded an initiative」而是「ran」、「built」或「launched」。
其他的破綻
除了用字之外，還要留意：
到處都是破折號。AI 很愛用，但大多數人不會去打它，這也是為什麼這整篇文章大致上都避開它。
每一條條列都用同樣的方式開頭、跑同樣的長度。真實的職涯是凹凸不平的，所以讓你的開頭動詞有變化。
殘留的佔位符。一份送出去的履歷裡留著「[在此加入數據]」，是個真實又常見的破綻。
文字裡隨機散落的粗體字。粗體屬於標題，不是拿來灑著強調用的。
你可以讓 ChatGPT 自己做掉大部分的清理工作：
Rewrite this resume text to sound like a real person wrote it. Vary the sentence length and the opening verbs. Remove buzzwords like spearheaded, leveraged, dynamic, and "at the intersection of." Do not use em dashes. Keep every fact exactly as written.
{{resume_text}}
然後再加回 AI 給不了的東西：那個具體的名詞。客戶的名字、確切的工具、團隊的人數、真實的專案。這些細節既是讓你可信的東西，也正是 ChatGPT 會漏掉的東西。如果你想要一套更深入的方法，把你自己的聲音保留在 AI 寫作裡，我們寫了一整篇關於讓 AI 提示詞聽起來像你的指南。
