The posting closes Friday. The resume on your laptop still says "responsible for" four times, and the one number you are proud of is buried in the third bullet. The popular lists of ChatGPT resume prompts run to two dozen variations of "make this sound professional" and rarely show what the model sent back.
This guide runs eight prompts on a real fact sheet, in the order you would use them, and shows what came back. A fictional candidate, Jordan Reyes, a marketing coordinator applying for a marketing manager role, stands in for you. Every output is quoted as the model returned it, including the places where it stretched the truth and the prompt that caught it.
Key Takeaways
- ChatGPT writes a usable resume from a fact sheet you prepare first: titles, dates, numbers, tools.
- Eight prompts cover the whole job: summary, bullets, tailoring to a posting, review, and a final honesty check.
- The tailoring prompt takes two turns: paste the posting and your resume, ask for the gap list, then rewrite.
- The same prompts run on GPT-6 Luna through GPT-6 on Atlas Cloud without a ChatGPT plan.
- One in five recruiters in a 2025 survey would reject an AI-generated resume, so every line must trace back to your facts.
Can ChatGPT Write a Resume? What It Needs From You First
Yes, and the result is decided before you type the first prompt. ChatGPT has no record of your career. Whatever is missing from the chat tends to get replaced with a plausible phrase, and plausible phrases are what recruiters skim past.
So the first document is a fact sheet, in plain lines:
- Each job: title, employer, kind of company, start and end month.
- One number per result: open rate 21% to 29%, cost per purchase $31 to $22, 48,000 subscribers.
- Tools by their real names: Klaviyo, Meta Ads Manager, Google Analytics 4, not "marketing platforms".
- The posting lines you actually meet, pasted as written.
- The role you are targeting, with the company name.
Jordan's sheet ran 22 lines. The infographic shows where each kind of line ends up.
Two things stay off the sheet. Your address, phone number and references' details add nothing to a draft; the ChatGPT cover letter guide has a full table of what to keep out of the chat box. And figures from your current employer go in only in the form you are allowed to share.
A note on versions, since searches for "ChatGPT 5.4 resume writing" are still common. GPT-5.4 reached ChatGPT in March 2026 and is no longer the default. OpenAI's help center describes GPT-6 as rolling out as the default in Chat for paid plans, and OpenAI's GPT-6 launch post says the Free and Go tiers get GPT-6 Luna; our GPT-6 availability guide tracks which tier gets which model. The outputs below came from a free account in a Temporary Chat in October 2026, and ChatGPT tagged all eight replies gpt-6.
8 ChatGPT Resume Prompts, in the Order You Use Them
Each prompt went into the same chat, after the fact sheet. Prompts 1, 2, 4 and 5 build the resume. Prompt 3 fixes weak bullets, and we ran it on GPT-6 Luna in the next section. Prompts 6, 7 and 8 tailor, review and verify, and get their own sections.
Prompt 1: the first draft
plaintext1Here is my fact sheet. Build a one-page resume from it with these sections: Summary, Experience, Skills, Education and Certifications. Use only facts from the sheet. Do not add any skill, number or responsibility that is not listed. Each Experience bullet starts with a verb and includes a number where the sheet gives one. Plain text, no tables. 2 3[paste the fact sheet]
ChatGPT returned a complete page with every number intact. It also put "Marketing Manager Candidate" under the name, a title nobody holds, and listed "Campaign Planning and Execution" as a skill, a category it inferred rather than a tool Jordan named. Prompt 8 catches the title; prompt 4 replaces the skills line.
Prompt 2: the summary
plaintext1Write a three-sentence professional summary for the top of this resume. Sentence one: my current title, how long I have done it, and the one result I am proudest of. Sentence two: the two channels or tools I use most. Sentence three: the role I am targeting. No adjectives like passionate, dynamic or results-driven. Use only facts from the sheet. Reply in the chat, not in a canvas.
Three sentences, one number, no adjectives. Give the model a slot for each fact and it has nowhere to put filler. The last line matters too: without it, ChatGPT had opened a canvas for the first draft, which is harder to copy from.
Prompt 3: weak bullets into strong ones
plaintext1Rewrite these three resume bullets. Each one should start with a strong verb, say what I did, and end with the measurable result. Keep every number exactly as given and add none. Maximum 25 words each. 2 3Bullets: 4[paste the bullets] 5 6Facts you may use: [the numbers behind each bullet]
Prompt 4: the skills section
plaintext1Write a Skills section from the fact sheet as three labeled lines: Platforms, Analytics and reporting, Content. List only tools named in the sheet. No soft skills.
Back came exactly three lines: Platforms (Klaviyo, Meta Ads Manager, Shopify admin, Asana), Analytics and reporting (Google Analytics 4, Looker Studio), Content (Canva, Adobe Premiere, marked basic). Same model and facts as prompt 1; this one names the labels and bans everything else.
Prompt 5: an objective for a step up
plaintext1Write a two-sentence resume objective for someone moving from coordinator to manager. Sentence one: what I have run and at what scale (budget, list size, people). Sentence two: what I want to own next. No phrases like "seeking a challenging position".
The reply led with the $120K budget, the 48,000-subscriber list and the three freelancers, then said Jordan wants "to lead marketing strategy and campaign execution". Strategy appears nowhere on the fact sheet. The objective is where ambition creeps in, which is why the honesty check re-reads it.
Run the Same Resume Prompts on GPT-6 in the Atlas Cloud Playground
The prompts are plain text, so they run on any GPT-6 model you can reach. Atlas Cloud serves the OpenAI family through one account, so you pick GPT-6 Luna or GPT-6 Sol by name instead of taking whatever your ChatGPT tier assigns, and an API request carries only the text you put in it. We sent prompt 3 to GPT-6 Luna with three deliberately weak bullets.
Method 1: Paste the Prompt in the GPT-6 Luna Playground
Sign in and open the GPT-6 Luna playground, then:
- Paste a prompt into the message box. The fact sheet goes in the same message as prompt 1; later prompts refer back to it in the same thread.
- Click the send arrow. The reply streams in, and the next prompt goes in the same box.
Method 2: Send the Resume Prompt Through the API
If you tailor one resume for a dozen postings, a script saves the retyping. Three steps.
Step 1: Create an API key. Open Settings in the console, go to API Keys, and click Create API Key. Keep it on the server side.
Step 2: Check the request format. The API docs cover the chat completions format, and the model page has an API tab with a ready-made example.
Step 3: Send the request. Prompt and fact sheet go in one user message.
plaintext1curl -X POST "https://api.atlascloud.ai/v1/chat/completions" \ 2 -H "Authorization: Bearer $ATLASCLOUD_API_KEY" \ 3 -H "Content-Type: application/json" \ 4 -d '{ 5 "model": "openai/gpt-6-luna", 6 "messages": [{"role": "user", "content": "Rewrite these three resume bullets. Each one should start with a strong verb, say what I did, and end with the measurable result. Keep every number exactly as given and add none. Maximum 25 words each. Bullets: ... Facts you may use: ..."}], 7 "max_tokens": 1500 8 }'
The text comes back in choices[0].message.content. The same endpoint and key serve the rest of the OpenAI family, so switching to GPT-6 Sol for a harder rewrite means changing only the model string.
The ChatGPT Prompt to Tailor Your Resume to a Job Description
Asking for the tailored rewrite in one step is where resumes pick up claims nobody made. Split it in two. The first turn only maps the posting against what you have.
Prompt 6, turn one: the gap table
plaintext1Here is the job posting. Compare it with the resume you built. Make a table with three columns: requirement from the posting, the line in my resume that answers it, and a status of Covered, Partly covered or Missing. Do not rewrite anything yet. 2 3[paste the posting]
ChatGPT produced fourteen rows for Northgate's posting. Three of them:
| Requirement from the posting | Line in the resume that answers it | Status |
|---|---|---|
| Manage a paid media budget of about $500K a year across Meta and Google | "Manage a $120K annual Meta paid social budget, reducing cost per purchase from $31 to $22 over 12 months." | Partly covered |
| Lead and develop a team of three | "Coordinate 2 freelance designers and 1 photographer, and build the content calendar in Asana." | Partly covered |
| Experience with Google Ads | No Google Ads experience is listed in the resume or fact sheet. | Missing |
The count at the bottom, 8 covered, 5 partly, 1 missing, tells you what to rewrite and what to leave alone before anything gets rewritten.
Prompt 6, turn two: rewrite only what is true
plaintext1Now rewrite only the bullets marked Partly covered so they use the posting's wording where it is true. For Missing items, do not invent anything; list them under "Not on the resume" and I will decide what to do.
The rewrite touched five bullets and put Google Ads under "Not on the resume". It did not conjure a $500K budget or a team of three. It did add two soft phrases: the Asana calendar bullet gained "coordinating marketing content across email, paid social, and in-store channels", and the freelancer bullet gained "to support marketing content production". Neither is false, exactly. Neither is on the fact sheet either.
How to Use ChatGPT to Improve a Resume You Already Have
If you have a resume from two years ago rather than a blank page, paste the old resume as the fact sheet, add lines for everything since, and run prompt 1 with "update this resume" in place of "build a resume". From there, prompts 6, 7 and 8 do the improving, and the two review prompts work best when you ask for a scorecard instead of an opinion.
ChatGPT Resume Review Prompt: Ask for a Scorecard, Not Praise
Prompt 7: the hiring manager's scorecard
plaintext1Review the tailored resume as the hiring manager for this posting. Score it 1 to 5 on five things: match to the requirements, evidence (a number in each bullet), clarity, length, and anything that reads as AI-written. For each score give one sentence and one specific fix. No praise sentences.
Scores of 3, 3, 4, 4 and 3. The model marked its own tailoring down for evidence, because the rewritten calendar and freelancer bullets had lost their numbers, and named "support marketing content production" as the phrase that reads as machine-made. It wrote that phrase two prompts earlier. A review prompt with a "no praise" clause will say so.
Prompt 8: the honesty check
plaintext1Compare the tailored resume against my original fact sheet. List every claim, number, tool or responsibility in the resume that does not appear in the fact sheet, quoting the exact phrase. If there are none, say "None".
Nine phrases: the invented title from prompt 1, "marketing strategy" from the objective, both soft phrases from the tailoring turn, and a subtle one, "build and maintain the content calendar", where the sheet only says Jordan built it. It also flagged "lead a team of three" and "improve conversion rate", the posting's own words, noting they would only matter if carried into the resume. Delete what you cannot back, keep the rest, and the draft is yours to send.
Is It Okay to Use ChatGPT for a Resume? What Hiring Managers Say
Using ChatGPT on a resume is common enough that the people reading resumes have formed views on it. The numbers are self-reported and each survey had its own sample, so read them as the shape of the room.
TopResume's survey of 600 US hiring managers, fielded in May 2025, found that "One in five (19.6%) recruiters would reject a candidate with an AI-generated resume or cover letter", and a third said they could spot one in under twenty seconds. Stated intent rather than a measured rejection rate, and still a fifth of the room.
Robert Half's survey of more than 2,000 US hiring managers, fielded in November 2025, found 67% of HR leaders saying AI-generated applications had slowed hiring and 65% saying candidate skills had become harder to verify. The complaint is about claims the reader cannot check.
On the applicant side, Novoresume's Resume Report 2026, a May 2026 survey of 2,000 US adults, put the share who used AI on their last resume at 42.6%, and found that people who let AI write most of the document were eight times more likely to have been confronted about it by a recruiter or employer. The tool is normal; the unbacked claim is what gets caught. Prompt 8 is the cheapest insurance against the second group.
Frequently Asked Questions
Is it safe to put your resume in ChatGPT?
The file is fine to upload through the + button in the message box, but strip the header first; an address, phone number and references add nothing to a draft. Use a Temporary Chat for the session. OpenAI's help center says a temporary chat is not used to improve its models while it stays temporary, and that a copy may be kept for up to 30 days for safety purposes.
Does ChatGPT make good resumes?
It makes resumes as good as the facts you give it. In our run, the first draft kept every number from the fact sheet and invented one job title; the tailored version added two soft phrases that a review prompt then flagged. A specific sheet gets a strong draft, a thin sheet gets a generic one.
Can ChatGPT make a resume template?
It returns a plain-text layout with section headings, which is what prompt 1 gives you, but the formatting happens wherever you paste the text, usually a Google Docs or Word template you already have. Keep the model on the words and let the document app handle the margins.
ChatGPT or Gemini for a resume?
Nothing in these prompts is specific to ChatGPT, so they paste into Gemini unchanged; we ran them only on GPT-6. Use whichever assistant you already use for everything else, and run prompt 8 at the end either way.
Which ChatGPT version is best for resume writing, GPT-6 or ChatGPT 5.4?
GPT-5.4 is the version people search for because it was the default in spring 2026; it has since been replaced. Every ChatGPT reply in this guide was tagged gpt-6 on a free account, which OpenAI's launch post says runs GPT-6 Luna. If your chat shows an older model, the prompts still run; repeat the honesty check.
How much does GPT-6 Luna cost on Atlas Cloud?
The model page lists GPT-6 Luna at $0.10 per million input tokens and $0.50 per million output tokens. A resume prompt with a fact sheet runs to roughly a thousand tokens each way, so one pass through all eight prompts costs well under a cent, and tailoring the same resume to twenty postings stays in the cents.
Conclusion
The resumes that get flagged tend to be the ones with a sentence nobody can trace back to a fact. Eight prompts kept Jordan's draft honest: a fact sheet first, a structured summary, bullets that end in a number, a two-turn tailoring pass that lists gaps before it rewrites, a scorecard review, and a last check that quotes every phrase the model added. The same prompts ran on GPT-6 in ChatGPT and on GPT-6 Luna in the Atlas Cloud playground, with the same discipline.
Keep the fact sheet, change the posting, and the whole sequence is eight pastes per application. That is what ChatGPT resume prompts are for: a repeatable way to make sure "results-driven" never appears on your resume again.
